WO2023091436A1 - Système et techniques de manipulation d'un texte long pour des modèles de langage préentraînés - Google Patents

Système et techniques de manipulation d'un texte long pour des modèles de langage préentraînés Download PDF

Info

Publication number
WO2023091436A1
WO2023091436A1 PCT/US2022/050024 US2022050024W WO2023091436A1 WO 2023091436 A1 WO2023091436 A1 WO 2023091436A1 US 2022050024 W US2022050024 W US 2022050024W WO 2023091436 A1 WO2023091436 A1 WO 2023091436A1
Authority
WO
WIPO (PCT)
Prior art keywords
token
chunk
utterances
utterance
skill
Prior art date
Application number
PCT/US2022/050024
Other languages
English (en)
Inventor
Thanh Tien Vu
Tuyen Quang Pham
Mark Edward Johnson
Thanh Long Duong
Ying Xu
Poorya Zaremoodi
Omid Mohamad NEZAMI
Budhaditya Saha
Cong Duy Vu Hoang
Original Assignee
Oracle International Corporation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US17/750,240 external-priority patent/US20230161963A1/en
Application filed by Oracle International Corporation filed Critical Oracle International Corporation
Publication of WO2023091436A1 publication Critical patent/WO2023091436A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G06F40/295—Named entity recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/10—Text processing
    • G06F40/12—Use of codes for handling textual entities
    • G06F40/131—Fragmentation of text files, e.g. creating reusable text-blocks; Linking to fragments, e.g. using XInclude; Namespaces

Definitions

  • An intelligent hot generally powered by artificial intelligence (Al), can communicate more intelligently and contextually in live conversations, and thus may allow for a more natural conversation between the hot and the end users for improved conversational experience. Instead of the end user learning a fixed set of keywords or commands that the hot knows how to respond to, an intelligent hot may be able to understand the end user’s intention based upon user utterances in natural language and respond accordingly.
  • Pre-trained language models used with natural language processing can be designed to support a maximum length of text.
  • the maximum length of text can be limited to 512 token pieces (e.g., 512 subwords).
  • Chabot entries that exceeded the maximum length of text would be truncated to less than the maximum length of text. This solution could result in loss of information from the truncated data and less accurate results. This leads to the reduced performance of the Chatbot and poor customer experience.
  • Techniques disclosed herein relate generally to catboats. More specifically and without limitation, techniques disclosed herein relate to techniques for handling long text for pre-trained language model.
  • pre-trained language models normally support a predetermined maximum number of token pieces.
  • Long text can be divided into overlapping chunks with predefined chunk size and number of overlapping token pieces.
  • the chunks can be tagged separately by the named entity recognition (NER).
  • NER named entity recognition
  • the two scores of annotated labels for each overlapping token pieces (one from the first chunk and another from the second chunk) can be combined to determine the final label for each token piece in the long text. This can reduce the training and inferencing time while maintaining high performance of the Chabot.
  • the dividing the set of utterances with a size N is broken into (N- L)/(K-L) overlapping chunks.
  • K is a chunk size
  • L is an overlapping size (K > L).
  • the determining the overall score and label of a token piece is based on a confidence score from the first selected chunk of the plurality of chunks.
  • the determining the overall score is based on a maximum confidence score of the plurality of chunks.
  • the predetermined number of token pieces (i. e. , the chunk size) is 512 tokens and the predetermined number of overlapping tokens is 128. In some aspects, the predetermined number of token pieces is 32 and the predetermined number of overlapping tokens is 8.
  • a system includes one or more data processors and a non-transitory computer readable storage medium containing instructions which, when executed on the one or more data processors, cause the one or more data processors to perform part or all of one or more methods disclosed herein.
  • a computer-program product is provided that is tangibly embodied in a non-transitory machine-readable storage medium and that includes instructions configured to cause one or more data processors to perform part or all of one or more methods disclosed herein.
  • FIG. 1 is a simplified block diagram of an environment incorporating an exemplary embodiment of a Chabot system according to certain aspects.
  • FIG. 2 is a simplified block diagram of a computing system implementing a master bot according to certain aspects.
  • FIG. 3 is a simplified block diagram of a computing system implementing a skill bot according to certain aspects.
  • FIG. 4 illustrates a system for handling long text for pre-trained language models.
  • FIG. 5 illustrates converting an exemplary utterance into a plurality of chunks.
  • FIG. 6 illustrates techniques for merging the predictions.
  • FIG. 7 is a flow diagram for a technique for handling long text for pre-trained language models.
  • FIG. 8 depicts a simplified diagram of a distributed system for implementing certain aspects.
  • FIG. 9 is a simplified block diagram of one or more components of a system environment by which services provided by one or more components of an embodiment system may be offered as cloud services, in accordance with certain aspects.
  • FIG. 10 illustrates an example computer system that may be used to implement certain aspects.
  • a bot (also referred to as a skill, Chabot, chatterbot, or talkbot) is a computer program that can perform conversations with end users.
  • the bot can generally respond to natural-language messages (e.g., questions or comments) through a messaging application that uses natural-language messages.
  • Enterprises may use one or more bot systems to communicate with end users through a messaging application.
  • the messaging application which may be referred to as a channel, may be an end user preferred messaging application that the end user has already installed and familiar with. Thus, the end user does not need to download and install new applications in order to chat with the bot system.
  • the messaging application may include, for example, over-the-top (OTT) messaging channels (such as Facebook Messenger, Facebook WhatsApp, WeChat, Line, Kik, Telegram, Talk, Skype, Slack, or Short Message Service (SMS), virtual private assistants (such as Amazon Dot, Echo, or Show, Google Home, Apple HomePod, etc.), mobile and web app extensions that extend native or hybrid/responsive mobile apps or web applications with chat capabilities, or voice based input (such as devices or apps with interfaces that use Siri, Cortana, Google Voice, or other speech input for interaction).
  • OTT over-the-top
  • SMS Short Message Service
  • virtual private assistants such as Amazon Dot, Echo, or Show, Google Home, Apple HomePod, etc.
  • mobile and web app extensions that extend native or hybrid/responsive mobile apps or web applications with chat capabilities
  • voice based input such as devices or apps with interfaces that use Siri, Cortana, Google Voice, or other speech input for interaction.
  • a hot system may be associated with a Uniform Resource Identifier (URI).
  • the URI may identify the hot system using a string of characters.
  • the URI may be used as a webhook for one or more messaging application systems.
  • the URI may include, for example, a Uniform Resource Locator (URL) or a Uniform Resource Name (URN).
  • the hot system may be designed to receive a message (e.g., a hypertext transfer protocol (HTTP) post call message) from a messaging application system.
  • HTTP post call message may be directed to the URI from the messaging application system.
  • the message may be different from a HTTP post call message.
  • the hot system may receive a message from a SMS. While discussion herein may refer to communications that the hot system receives as a message, it should be understood that the message may be an HTTP post call message, a SMS message, or any other type of communication between two systems.
  • End users may interact with the hot system through a conversational interaction (sometimes referred to as a conversational user interface (UI)), just as interactions between people.
  • the interaction may include the end user saying “Hello” to the hot and the hot responding with a “Hi” and asking the end user how it can help.
  • the interaction may also be a transactional interaction with, for example, a banking bot, such as transferring money from one account to another; an informational interaction with, for example, a HR bot, such as checking for vacation balance; or an interaction with, for example, a retail bot, such as discussing returning purchased goods or seeking technical support.
  • the bot system may intelligently handle end user interactions without interaction with an administrator or developer of the bot system. For example, an end user may send one or more messages to the bot system in order to achieve a desired goal.
  • a message may include certain content, such as text, emojis, audio, image, video, or other method of conveying a message.
  • the bot system may convert the content into a standardized form (e.g., a representational state transfer (REST) call against enterprise services with the proper parameters) and generate a natural language response.
  • the hot system may also prompt the end user for additional input parameters or request other additional information.
  • the hot system may also initiate communication with the end user, rather than passively responding to end user utterances.
  • Described herein are various techniques for identifying an explicit invocation of a hot system and determining an input for the hot system being invoked.
  • explicit invocation analysis is performed by a master bot based on detecting an invocation name in an utterance.
  • the utterance may be refined for input to a skill bot associated with the invocation name.
  • a conversation with a bot may follow a specific conversation flow including multiple states.
  • the flow may define what would happen next based on an input.
  • a state machine that includes user defined states (e.g., end user intents) and actions to take in the states or from state to state may be used to implement the bot system.
  • a conversation may take different paths based on the end user input, which may impact the decision the bot makes for the flow. For example, at each state, based on the end user input or utterances, the bot may determine the end user’s intent in order to determine the appropriate next action to take.
  • intent refers to an intent of the user who provided the utterance.
  • the user may intend to engage a bot in conversation for ordering pizza, so that the user’s intent could be represented through the utterance “Order pizza.”
  • a user intent can be directed to a particular task that the user wishes a Chabot to perform on behalf of the user. Therefore, utterances can be phrased as questions, commands, requests, and the like, that reflect the user’s intent.
  • An intent may include a goal that the end user would like to accomplish.
  • the term “intent” is used herein to refer to configuration information for mapping a user’s utterance to a specific task/action or category of task/action that the Chabot can perform.
  • intent In order to distinguish between the intent of an utterance (i.e. , a user intent) and the intent of a Chabot, the latter is sometimes referred to herein as a “bot intent.”
  • a bot intent may comprise a set of one or more utterances associated with the intent. For instance, an intent for ordering pizza can have various permutations of utterances that express a desire to place an order for pizza.
  • a bot intent may be associated with one or more dialog flows for starting a conversation with the user and in a certain state.
  • the first message for the order pizza intent could be the question “What kind of pizza would you like?”
  • a hot intent may further comprise named entities that relate to the intent.
  • the order pizza intent could include variables or parameters used to perform the task of ordering pizza, e.g., topping 1, topping 2, pizza type, pizza size, pizza quantity, and the like. The value of an entity is typically obtained through conversing with the user.
  • FIG. 1 is a simplified block diagram of an environment 100 incorporating a Chabot system according to certain aspects.
  • Environment 100 comprises a digital assistant builder platform (DABP) 102 that enables users of DABP 102 to create and deploy digital assistants or Chabot systems.
  • DABP 102 can be used to create one or more digital assistants (or DAs) or Chabot systems.
  • user 104 representing a particular enterprise can use DABP 102 to create and deploy a digital assistant 106 for users of the particular enterprise.
  • DABP 102 can be used by a bank to create one or more digital assistants for use by the bank's customers.
  • the same DABP 102 platform can be used by multiple enterprises to create digital assistants.
  • an owner of a restaurant e.g., a pizza shop
  • a "digital assistant” is an entity that helps users of the digital assistant accomplish various tasks through natural language conversations.
  • a digital assistant can be implemented using software only (e.g., the digital assistant is a digital entity implemented using programs, code, or instructions executable by one or more processors), using hardware, or using a combination of hardware and software.
  • a digital assistant can be embodied or implemented in various physical systems or devices, such as in a computer, a mobile phone, a watch, an appliance, a vehicle, and the like.
  • a digital assistant is also sometimes referred to as a Chabot system. Accordingly, for purposes of this disclosure, the terms digital assistant and Chabot system are interchangeable.
  • a digital assistant such as digital assistant 106 built using DABP 102, can be used to perform various tasks via natural language-based conversations between the digital assistant and its users 108.
  • a user may provide one or more user inputs 110 to digital assistant 106 and get responses 112 back from digital assistant 106.
  • a conversation can include one or more of inputs 110 and responses 112.
  • a user can request one or more tasks to be performed by the digital assistant and, in response, the digital assistant is configured to perform the user-requested tasks and respond with appropriate responses to the user.
  • User inputs are generally in a natural language form and are referred to as utterances.
  • a user utterance 110 can be in text form, such as when a user types in a sentence, a question, a text fragment, or even a single word and provides it as input to digital assistant 106.
  • a user utterance 110 can be in audio input or speech form, such as when a user says or speaks something that is provided as input to digital assistant 106.
  • the utterances are typically in a language spoken by the user 108. For example, the utterances may be in English, or some other language.
  • the speech input is converted to text form utterances in that particular language and the text utterances are then processed by digital assistant 106.
  • Various speech-to-text processing techniques may be used to convert a speech or audio input to a text utterance, which is then processed by digital assistant 106.
  • the speech-to-text conversion may be done by digital assistant 106 itself.
  • An utterance which may be a text utterance or a speech utterance, can be a fragment, a sentence, multiple sentences, one or more words, one or more questions, combinations of the aforementioned types, and the like.
  • Digital assistant 106 is configured to apply natural language understanding (NLU) techniques to the utterance to understand the meaning of the user input. As part of the NLU processing for an utterance, digital assistant 106 is configured to perform processing to understand the meaning of the utterance, which involves identifying one or more intents and one or more entities corresponding to the utterance. Upon understanding the meaning of an utterance, digital assistant 106 may perform one or more actions or operations responsive to the understood meaning or intents.
  • NLU natural language understanding
  • the utterances are text utterances that have been provided directly by a user 108 of digital assistant 106 or are the results of conversion of input speech utterances to text form. This however is not intended to be limiting or restrictive in any manner.
  • a user 108 input may request a pizza to be ordered by providing an utterance such as "I want to order a pizza.”
  • digital assistant 106 Upon receiving such an utterance, digital assistant 106 is configured to understand the meaning of the utterance and take appropriate actions. The appropriate actions may involve, for example, responding to the user with questions requesting user input on the type of pizza the user desires to order, the size of the pizza, any toppings for the pizza, and the like.
  • the responses provided by digital assistant 106 may also be in natural language form and typically in the same language as the input utterance. As part of generating these responses, digital assistant 106 may perform natural language generation (NLG).
  • the digital assistant may guide the user to provide all the requisite information for the pizza order, and then at the end of the conversation cause the pizza to be ordered.
  • Digital assistant 106 may end the conversation by outputting information to the user indicating that the pizza has been ordered.
  • digital assistant 106 performs various processing in response to an utterance received from a user.
  • this processing involves a series or pipeline of processing steps including, for example, understanding the meaning of the input utterance (sometimes referred to as Natural Language Understanding (NLU), determining an action to be performed in response to the utterance, where appropriate causing the action to be performed, generating a response to be output to the user responsive to the user utterance, outputting the response to the user, and the like.
  • NLU processing can include parsing the received input utterance to understand the structure and meaning of the utterance, refining and reforming the utterance to develop a better understandable form (e.g., logical form) or structure for the utterance.
  • Generating a response may include using NLG techniques.
  • the NLU processing performed by a digital assistant can include various NLP related processing such as sentence parsing (e.g., tokenizing, lemmatizing, identifying part-of-speech tags for the sentence, identifying named entities in the sentence, generating dependency trees to represent the sentence structure, splitting a sentence into clauses, analyzing individual clauses, resolving anaphoras, performing chunking, and the like).
  • sentence parsing e.g., tokenizing, lemmatizing, identifying part-of-speech tags for the sentence, identifying named entities in the sentence, generating dependency trees to represent the sentence structure, splitting a sentence into clauses, analyzing individual clauses, resolving anaphoras, performing chunking, and the like.
  • the NLU processing or portions thereof is performed by digital assistant 106 itself.
  • digital assistant 106 may use other resources to perform portions of the NLU processing.
  • the syntax and structure of an input utterance sentence may be identified by processing the sentence using a parser, a part-of-speech tagger, and/or a named entity recognizer.
  • a parser, a part-of-speech tagger, and a named entity recognizer such as ones provided by the Stanford Natural Language Processing (NLP) Group are used for analyzing the sentence structure and syntax. These are provided as part of the Stanford CoreNLP toolkit.
  • NLP Stanford Natural Language Processing
  • Digital assistant 106 may provide subsystems (e.g., components implementing NLU functionality) that are configured for performing processing for different languages. These subsystems may be implemented as pluggable units that can be called using service calls from an NLU core server. This makes the NLU processing flexible and extensible for each language, including allowing different orders of processing.
  • a language pack may be provided for individual languages, where a language pack can register a list of subsystems that can be served from the NLU core server.
  • a digital assistant such as digital assistant 106 depicted in FIG. 1, can be made available or accessible to its users 108 through a variety of different channels, such as but not limited to, via certain applications, via social media platforms, via various messaging services and applications, and other applications or channels.
  • a single digital assistant can have several channels configured for it so that it can be run on and be accessed by different services simultaneously.
  • a digital assistant or Chabot system generally contains or is associated with one or more skills.
  • these skills are individual catboats (referred to as skill bots) that are configured to interact with users and fulfill specific types of tasks, such as tracking inventory, submitting timecards, creating expense reports, ordering food, checking a bank account, making reservations, buying a widget, and the like.
  • digital assistant 106 or Chabot system includes first skills bot 116-1, second skills bot 116-2, and so on.
  • the terms “skill” and “skills” are used synonymously with the terms “skill bot” and “skill bots,” respectively.
  • Each skill associated with a digital assistant helps a user of the digital assistant complete a task through a conversation with the user, where the conversation can include a combination of text or audio inputs provided by the user and responses provided by the skill bots. These responses may be in the form of text or audio messages to the user and/or using simple user interface elements (e.g., select lists) that are presented to the user for the user to make selections.
  • simple user interface elements e.g., select lists
  • a skill or skill bot can be associated or added to a digital assistant.
  • a skill bot can be developed by an enterprise and then added to a digital assistant using DABP 102.
  • a skill hot can be developed and created using DABP 102 and then added to a digital assistant created using DABP 102.
  • DABP 102 provides an online digital store (referred to as a "skills store") that offers multiple skills directed to a wide range of tasks. The skills offered through the skills store may also expose various cloud services.
  • a user of DABP 102 can access the skills store via DABP 102, select a desired skill, and indicate that the selected skill is to be added to the digital assistant created using DABP 102.
  • a skill from the skills store can be added to a digital assistant as is or in a modified form (for example, a user of DABP 102 may select and clone a particular skill bot provided by the skills store, make customizations or modifications to the selected skill bot, and then add the modified skill bot to a digital assistant created using DABP 102).
  • digital assistants created and deployed using DABP 102 may be implemented using a master bot/child(or sub) bot paradigm or architecture.
  • a digital assistant is implemented as a master bot that interacts with one or more child hots that are skill hots.
  • digital assistant 106 comprises a master bot 114 and first skills bot 116-1, second skills botl 16-2, etc. that are child hots of master bot 114.
  • digital assistant 106 is itself considered to act as the master bot.
  • a digital assistant implemented according to the master-child bot architecture enables users of the digital assistant to interact with multiple skills through a unified user interface, namely via the master bot.
  • the user input is received by the master bot.
  • the master bot then performs processing to determine the meaning of the user input utterance.
  • the master bot determines whether the task requested by the user in the utterance can be handled by the master bot itself, else the master bot selects an appropriate skill bot for handling the user request and routes the conversation to the selected skill bot. This enables a user to converse with the digital assistant through a common single interface and still provide the capability to use several skill bots configured to perform specific tasks.
  • the master bot of the digital assistant may interface with skill bots with specific functionalities, such as a CRM bot for performing functions related to customer relationship management (CRM), an ERP bot for performing functions related to enterprise resource planning (ERP), an HCM bot for performing functions related to human capital management (HCM), etc.
  • CRM bot for performing functions related to customer relationship management
  • ERP bot for performing functions related to enterprise resource planning
  • HCM bot for performing functions related to human capital management
  • the master bot in a master hot/ child hots infrastructure, is configured to be aware of the available list of skill bots.
  • the master bot may have access to metadata that identifies the various available skill bots, and for each skill bot, the capabilities of the skill bot including the tasks that can be performed by the skill bot.
  • the master bot Upon receiving a user request in the form of an utterance, the master bot is configured to, from the multiple available skill bots, identify or predict a specific skill bot that can best serve or handle the user request.
  • the master bot then routes the utterance (or a portion of the utterance) to that specific skill bot for further handling. Control thus flows from the master bot to the skill bots.
  • the master bot can support multiple input and output channels.
  • routing may be performed with the aid of processing performed by one or more available skill bots.
  • a skill bot can be trained to infer an intent for an utterance and to determine whether the inferred intent matches an intent with which the skill bot is configured.
  • the routing performed by the master bot can involve the skill bot communicating to the master bot an indication of whether the skill bot has been configured with an intent suitable for handling the utterance.
  • digital assistant 106 comprising a master bot 114 and first skills bot 116-1, second skills bot 116-2, and third skills bot 116-3
  • a digital assistant can include various other components (e.g., other systems and subsystems) that provide the functionalities of the digital assistant.
  • These systems and subsystems may be implemented only in software (e.g., code, instructions stored on a computer-readable medium and executable by one or more processors), in hardware only, or in implementations that use a combination of software and hardware.
  • DABP 102 provides an infrastructure and various services and features that enable a user of DABP 102 to create a digital assistant including one or more skill bots associated with the digital assistant.
  • a skill bot can be created by cloning an existing skill bot, for example, cloning a skill bot provided by the skills store.
  • DABP 102 provides a skills store or skills catalog that offers multiple skill bots for performing various tasks.
  • a user of DABP 102 can clone a skill bot from the skills store. As needed, modifications or customizations may be made to the cloned skill hot.
  • a user of DABP 102 created a skill hot from scratch using tools and services offered by DABP 102.
  • the skills store or skills catalog provided by DABP 102 may offer multiple skill bots for performing various tasks.
  • creating or customizing a skill hot involves the following steps:
  • a skill bot designer can specify one or more invocation names for the skill bot being created. These invocation names can then be used by users of a digital assistant to explicitly invoke the skill bot. For example, a user can input an invocation name in the user's utterance to explicitly invoke the corresponding skill bot.
  • the skill bot designer specifies one or more intents (also referred to as bot intents) for a skill bot being created. The skill bot is then trained based upon these specified intents. These intents represent categories or classes that the skill bot is trained to infer for input utterances. Upon receiving an utterance, a trained skill bot infers an intent for the utterance, where the inferred intent is selected from the predefined set of intents used to train the skill bot. The skill bot then takes an appropriate action responsive to an utterance based upon the intent inferred for that utterance.
  • the intents for a skill bot represent tasks that the skill bot can perform for users of the digital assistant. Each intent is given an intent identifier or intent name. For example, for a skill bot trained for a bank, the intents specified for the skill hot may include “CheckBalance,” “TransferMoney,” “DepositCheck,” and the like.
  • the skill hot designer may also provide one or more example utterances that are representative of and illustrate the intent. These example utterances are meant to represent utterances that a user may input to the skill hot for that intent.
  • example utterances may include "Whaf s my savings account balance?”, "How much is in my checking account?", "How much money do I have in my account,” and the like. Accordingly, various permutations of typical user utterances may be specified as example utterances for an intent.
  • the intents and the their associated example utterances are used as training data to train the skill hot.
  • Various different training techniques may be used.
  • a predictive model is generated that is configured to take an utterance as input and output an intent inferred for the utterance by the predictive model.
  • input utterances are provided to an intent analysis engine, which is configured to use the trained model to predict or infer an intent for the input utterance.
  • the skill hot may then take one or more actions based upon the inferred intent.
  • the value associated with the AccountType entity is different for the two utterances.
  • This enables the skill bot to perform possibly different actions for the two utterances in spite of them resolving to the same intent.
  • One or more entities can be specified for certain intents configured for the skill bot. Entities are thus used to add context to the intent itself. Entities help describe an intent more fully and enable the skill bot to complete a user request.
  • built-in entities include, without limitation, entities related to time, date, addresses, numbers, email addresses, duration, recurring time periods, currencies, phone numbers, URLs, and the like.
  • Custom entities are used for more customized applications.
  • an AccountType entity may be defined by the skill bot designer that enables various banking transactions by checking the user input for keywords like checking, savings, and credit cards, etc.
  • a skill bot is configured to receive user input in the form of utterances parse or otherwise process the received input, and identify or select an intent that is relevant to the received user input. As indicated above, the skill bot has to be trained for this. In certain aspects, a skill bot is trained based upon the intents configured for the skill bot and the example utterances associated with the intents (collectively, the training data), so that the skill bot can resolve user input utterances to one of its configured intents. In certain aspects, the skill bot uses a predictive model that is trained using the training data and allows the skill bot to discern what users say (or in some cases, are trying to say).
  • DABP 102 provides various different training techniques that can be used by a skill bot designer to train a skill bot, including various machine-learning based training techniques, rules-based training techniques, and/or combinations thereof.
  • a portion (e.g., 80%) of the training data is used to train a skill bot model and another portion (e.g., the remaining 20%) is used to test or verify the model.
  • the trained model also sometimes referred to as the trained skill bot
  • a user's utterance may be a question that requires only a single answer and no further conversation.
  • a Q&A (question-and-answer) intent may be defined for a skill bot. This enables a skill bot to output replies to user requests without having to update the dialog definition.
  • Q&A intents are created in a similar manner as regular intents. The dialog flow for Q&A intents can be different from that for regular intents.
  • a dialog flow specified for a skill bot describes how the skill bot reacts as different intents for the skill bot are resolved responsive to received user input.
  • the dialog flow defines operations or actions that a skill bot will take, e.g., how the skill bot responds to user utterances, how the skill bot prompts users for input, how the skill hot returns data.
  • a dialog flow is like a flowchart that is followed by the skill hot.
  • the skill hot designer specifies a dialog flow using a language, such as markdown language.
  • a version of YAML called OBotML may be used to specify a dialog flow for a skill bot.
  • the dialog flow definition for a skill hot acts as a model for the conversation itself, one that lets the skill bot designer choreograph the interactions between a skill bot and the users that the skill bot services.
  • dialog flow definition for a skill bot contains three sections:
  • Context section The skill bot designer can define variables that are used in a conversation flow in the context section.
  • Other variables that may be named in the context section include, without limitation: variables for error handling, variables for built-in or custom entities, user variables that enable the skill bot to recognize and persist user preferences, and the like.
  • Default transitions section - Transitions for a skill bot can be defined in the dialog flow states section or in the default transitions section.
  • the transitions defined in the default transition section act as a fallback and get triggered when there are no applicable transitions defined within a state, or the conditions required to trigger a state transition cannot be met.
  • the default transitions section can be used to define routing that allows the skill bot to gracefully handle unexpected user actions.
  • States section - A dialog flow and its related operations are defined as a sequence of transitory states, which manage the logic within the dialog flow.
  • Each state node within a dialog flow definition names a component that provides the functionality needed at that point in the dialog. States are thus built around the components.
  • a state contains componentspecific properties and defines the transitions to other states that get triggered after the component executes.
  • Special case scenarios may be handled using the states sections. For example, there might be times when you want to provide users the option to temporarily leave a first skill they are engaged with to do something in a second skill within the digital assistant. For example, if a user is engaged in a conversation with a shopping skill (e.g., the user has made some selections for purchase), the user may want to jump to a banking skill (e.g., the user may want to ensure that he/she has enough money for the purchase), and then return to the shopping skill to complete the user's order. To address this, an action in the first skill can be configured to initiate an interaction with the second different skill in the same digital assistant and then return to the original flow.
  • a banking skill e.g., the user may want to ensure that he/she has enough money for the purchase
  • DABP 102 provides a set of preconfigured components for performing a wide range of functions. A skill bot designer can select one of more of these preconfigured components and associate them with states in the dialog flow for a skill bot. The skill bot designer can also create custom or new components using tools provided by DABP 102 and associate the custom components with one or more states in the dialog flow for a skill bot.
  • Testing and deploying the skill bot - DABP 102 provides several features that enable the skill bot designer to test a skill bot being developed. The skill bot can then be deployed and included in a digital assistant.
  • built-in system intents may be configured for the digital assistant. These built- in system intents are used to identify general tasks that the digital assistant itself (i.e., the master bot) can handle without invoking a skill bot associated with the digital assistant. Examples of system intents defined for a master bot include: (1) Exit: applies when the user signals the desire to exit the current conversation or context in the digital assistant; (2) Help: applies when the user asks for help or orientation; and (3) Unresolvedintent: applies to user input that doesn't match well with the exit and help intents.
  • the digital assistant also stores information about the one or more skill hots associated with the digital assistant. This information enables the master bot to select a particular skill bot for handling an utterance.
  • the digital assistant when a user inputs a phrase or utterance to the digital assistant, the digital assistant is configured to perform processing to determine how to route the utterance and the related conversation.
  • the digital assistant determines this using a routing model, which can be rules-based, Al-based, or a combination thereof.
  • the digital assistant uses the routing model to determine whether the conversation corresponding to the user input utterance is to be routed to a particular skill for handling, is to be handled by the digital assistant or master hot itself per a built-in system intent or is to be handled as a different state in a current conversation flow.
  • the digital assistant determines if the user input utterance explicitly identifies a skill hot using its invocation name. If an invocation name is present in the user input, then it is treated as explicit invocation of the skill hot corresponding to the invocation name. In such a scenario, the digital assistant may route the user input to the explicitly invoked skill hot for further handling. If there is no specific or explicit invocation, in certain aspects, the digital assistant evaluates the received user input utterance and computes confidence scores for the system intents and the skill hots associated with the digital assistant. The score computed for a skill hot or system intent represents how likely the user input is representative of a task that the skill hot is configured to perform or is representative of a system intent.
  • Any system intent or skill hot with an associated computed confidence score exceeding a threshold value is selected as a candidate for further evaluation.
  • the digital assistant selects, from the identified candidates, a particular system intent or a skill hot for further handling of the user input utterance.
  • the intents associated with those candidate skills are evaluated (according to the intent model for each skill) and confidence scores are determined for each intent.
  • a threshold value e.g. 70%
  • the user utterance is routed to that skill hot for further processing.
  • a system intent is selected, then one or more actions are performed by the master hot itself according to the selected system intent.
  • FIG. 2 is a simplified block diagram of a master bot (MB) system 200 according to certain aspects.
  • MB system 200 can be implemented in software only, hardware only, or a combination of hardware and software.
  • MB system 200 includes a pre-processing subsystem 210, a multiple intent subsystem (MIS) 220, an explicit invocation subsystem (EIS) 230, a skill bot invoker 240, and a data store 250.
  • MIS multiple intent subsystem
  • EIS explicit invocation subsystem
  • a skill bot invoker 240 a skill bot invoker 240
  • a data store 250 a data store 250.
  • MB system 200 depicted in FIG. 2 is merely an example of an arrangement of components in a master bot.
  • MB system 200 may have more or fewer systems or components than those shown in FIG. 2, may combine two or more subsystems, or may have a different configuration or arrangement of subsystems.
  • Pre-processing subsystem 210 receives an utterance “A” 202 from a user and processes the utterance through a language detector 212 and a language parser 214.
  • an utterance can be provided in various ways including audio or text.
  • the utterance 202 can be a sentence fragment, a complete sentence, multiple sentences, and the like.
  • Utterance 202 can include punctuation.
  • the pre-processing subsystem 210 may convert the audio to text using a speech-to-text converter (not shown) that inserts punctuation marks into the resulting text, e.g., commas, semicolons, periods, etc.
  • Language detector 212 detects the language of the utterance 202 based on the text of the utterance 202. The manner in which the utterance 202 is handled depends on the language since each language has its own grammar and semantics. Differences between languages are taken into consideration when analyzing the syntax and structure of an utterance.
  • Language parser 214 parses the utterance 202 to extract part of speech (POS) tags for individual linguistic units (e.g., words) in the utterance 202.
  • POS tags include, for example, noun (NN), pronoun (PN), verb (VB), and the like.
  • Language parser 214 may also tokenize the linguistic units of the utterance 202 (e.g., to convert each word into a separate token) and lemmatize words.
  • a lemma is the main form of a set of words as represented in a dictionary (e.g., “run” is the lemma for run, runs, ran, running, etc.).
  • the language parser 214 can perform other types of preprocessing that the language parser 214 can perform include chunking of compound expressions, e.g., combining “credit” and “card” into a single expression “credit_card.”
  • Language parser 214 may also identify relationships between the words in the utterance 202. For example, in some aspects, the language parser 214 generates a dependency tree that indicates which part of the utterance (e.g., a particular noun) is a direct object, which part of the utterance is a preposition, and so on.
  • the results of the processing performed by the language parser 214 form extracted information 205 and are provided as input to MIS 220 together with the utterance 202 itself.
  • the pre-processing system 210 can include a named entity recognizer 216 can be used to recognize certain utterances 202 or portions thereof.
  • the utterance 202 can include more than one sentence.
  • the utterance 202 can be treated as a single unit even if it includes multiple sentences.
  • pre-processing can be performed, e.g., by the pre-processing subsystem 210, to identify a single sentence among multiple sentences for multiple intents analysis and explicit invocation analysis.
  • the results produced by MIS 220 and EIS 230 are substantially the same regardless of whether the utterance 202 is processed at the level of an individual sentence or as a single unit comprising multiple sentences.
  • MIS 220 determines whether the utterance 202 represents multiple intents. Although MIS 220 can detect the presence of multiple intents in the utterance 202, the processing performed by MIS 220 does not involve determining whether the intents of the utterance 202 match to any intents that have been configured for a bot. Instead, processing to determine whether an intent of the utterance 202 matches a bot intent can be performed by an intent classifier 242 of the MB system 200 or by an intent classifier of a skill bot (e.g., as shown in the embodiment of FIG. 3). The processing performed by MIS 220 assumes that there exists a bot (e.g., a particular skill bot or the master bot itself) that can handle the utterance 202. Therefore, the processing performed by MIS 220 does not require knowledge of what hots are in the Chabot system (e.g., the identities of skill hots registered with the master bot), or knowledge of what intents have been configured for a particular bot.
  • a bot e.g., a particular
  • the MIS 220 applies one or more rules from a set of rules 252 in the data store 250.
  • the rules applied to the utterance 202 depend on the language of the utterance 202 and may include sentence patterns that indicate the presence of multiple intents.
  • a sentence pattern may include a coordinating conjunction that joins two parts (e.g., conjuncts) of a sentence, where both parts correspond to a separate intent. If the utterance 202 matches the sentence pattern, it can be inferred that the utterance 202 represents multiple intents.
  • an utterance with multiple intents does not necessarily have different intents (e.g., intents directed to different hots or to different intents within the same bot). Instead, the utterance could have separate instances of the same intent, e.g., “Place a pizza order using payment account X, then place a pizza order using payment account Y.”
  • the MIS 220 determines what portions of the utterance 202 are associated with each intent. MIS 220 constructs, for each intent represented in an utterance containing multiple intents, a new utterance for separate processing in place of the original utterance, e.g., an utterance “B” 206 and an utterance “C” 208, as depicted in FIG. 2. Thus, the original utterance 202 can be split into two or more separate utterances that are handled one at a time. MIS 220 determines, using the extracted information 205 and/or from analysis of the utterance 202 itself, which of the two or more utterances should be handled first.
  • MIS 220 may determine that the utterance 202 contains a marker word indicating that a particular intent should be handled first.
  • the newly formed utterance corresponding to this particular intent e.g., one of utterance 206 or utterance 208 will be the first to be sent for further processing by EIS 230.
  • the next highest priority utterance e.g., the other one of utterance 206 or utterance 208 can then be sent to the EIS 230 for processing.
  • EIS 230 determines whether the utterance that it receives (e.g., utterance 206 or utterance 208) contains an invocation name of a skill hot.
  • each skill hot in a Chabot system is assigned a unique invocation name that distinguishes the skill hot from other skill hots in the Chabot system.
  • a list of invocation names can be maintained as part of skill bot information 254 in data store 250. An utterance is deemed to be an explicit invocation when the utterance contains a word match to an invocation name.
  • the utterance received by the EIS 230 is deemed a non-explicitly invoking utterance 234 and is input to an intent classifier (e.g., intent classifier 242) of the master bot to determine which bot to use for handling the utterance.
  • intent classifier 242 will determine that the master bot should handle a non-explicitly invoking utterance. In other instances, the intent classifier 242 will determine a skill bot to route the utterance to for handling.
  • the explicit invocation functionality provided by the EIS 230 has several advantages. It can reduce the amount of processing that the master bot has to perform. For example, when there is an explicit invocation, the master bot may not have to do any intent classification analysis (e.g., using the intent classifier 242), or may have to do reduced intent classification analysis for selecting a skill bot. Thus, explicit invocation analysis may enable selection of a particular skill bot without resorting to intent classification analysis.
  • intent classification analysis e.g., using the intent classifier 242
  • explicit invocation analysis may enable selection of a particular skill bot without resorting to intent classification analysis.
  • EIS 230 may reformat the part to be sent to the invoked hot, e.g., to form a complete sentence.
  • the EIS 230 determines not only that there is an explicit invocation, but also what to send to the skill hot when there is an explicit invocation. In some instances, there may not be any text to input to the hot being invoked.
  • the EIS 230 could determine that the pizza hot is being invoked, but there is no text to be processed by the pizza hot. In such scenarios, the EIS 230 may indicate to the skill hot invoker 240 that there is nothing to send.
  • Skill hot invoker 240 invokes a skill hot in various ways. For instance, skill hot invoker 240 can invoke a hot in response to receiving an indication 235 that a particular skill hot has been selected as a result of an explicit invocation. The indication 235 can be sent by the EIS 230 together with the input for the explicitly invoked skill hot. In this scenario, the skill hot invoker 240 will turn control of the conversation over to the explicitly invoked skill hot. The explicitly invoked skill hot will determine an appropriate response to the input from the EIS 230 by treating the input as a stand-alone utterance. For example, the response could be to perform a specific action or to start a new conversation in a particular state, where the initial state of the new conversation depends on the input sent from the EIS 230.
  • skill hot invoker 240 can invoke a skill hot through implicit invocation using the intent classifier 242.
  • the intent classifier 242 can be trained, using machine-learning and/or rules-based training techniques, to determine a likelihood that an utterance is representative of a task that a particular skill hot is configured to perform.
  • the intent classifier 242 is trained on different classes, one class for each skill hot. For instance, whenever a new skill hot is registered with the master hot, a list of example utterances associated with the new skill hot can be used to train the intent classifier 242 to determine a likelihood that a particular utterance is representative of a task that the new skill hot can perform.
  • the parameters produced as result of this training e.g., a set of values for parameters of a machine-learning model
  • the intent classifier 242 is implemented using a machine-learning model, as described in further detail herein.
  • Training of the machine-learning model may involve inputting at least a subset of utterances from the example utterances associated with various skill hots to generate, as an output of the machine-learning model, inferences as to which bot is the correct bot for handling any particular training utterance.
  • an indication of the correct bot to use for the training utterance may be provided as ground truth information.
  • the behavior of the machine-learning model can then be adapted (e.g., through back-propagation) to minimize the difference between the generated inferences and the ground truth information.
  • the confidence score in addition to meeting a threshold confidence score value, the confidence score must exceed the next highest confidence score by a certain win margin. Imposing such a condition would enable routing to a particular skill bot when the confidence scores of multiple skill hots each exceed the threshold confidence score value.
  • the skill bot invoker 240 hands over processing to the identified bot.
  • the identified bot is the master bot. Otherwise, the identified bot is a skill bot.
  • the skill bot invoker 240 will determine what to provide as input 247 for the identified bot.
  • the input 247 can be based on a part of an utterance that is not associated with the invocation, or the input 247 can be nothing (e.g., an empty string).
  • the input 247 can be the entire utterance.
  • Data store 250 comprises one or more computing devices that store data used by the various subsystems of the master hot system 200.
  • the data store 250 includes rules 252 and skill hot information 254.
  • the rules 252 include, for example, rules for determining, by MIS 220, when an utterance represents multiple intents and how to split an utterance that represents multiple intents.
  • the rules 252 further include rules for determining, by EIS 230, which parts of an utterance that explicitly invokes a skill hot to send to the skill hot.
  • the skill hot information 254 includes invocation names of skill hots in the Chabot system, e.g., a list of the invocation names of all skill hots registered with a particular master hot.
  • the skill hot information 254 can also include information used by intent classifier 242 to determine a confidence score for each skill hot in the Chabot system, e.g., parameters of a machine-learning model.
  • FIG. 3 is a simplified block diagram of a skill bot system 300 according to certain aspects.
  • Skill bot system 300 is a computing system that can be implemented in software only, hardware only, or a combination of hardware and software. In certain aspects such as the embodiment depicted in FIG. 1, skill bot system 300 can be used to implement one or more skill bots within a digital assistant.
  • Intent classifier 320 can be trained in a similar manner to the intent classifier 242 discussed above in connection with the embodiment of FIG. 2 and as described in further detail herein.
  • the intent classifier 320 is implemented using a machine-learning model.
  • the machine-learning model of the intent classifier 320 is trained for a particular skill hot, using at least a subset of example utterances associated with that particular skill hot as training utterances.
  • the ground truth for each training utterance would be the particular hot intent associated with the training utterance.
  • the utterance 302 can be received directly from the user or supplied through a master hot.
  • a master hot e.g., as a result of processing through MIS 220 and EIS 230 in the embodiment depicted in FIG. 2, the MIS 310 can be bypassed so as to avoid repeating processing already performed by MIS 220.
  • MIS 310 can process the utterance 302 to determine whether the utterance 302 represents multiple intents.
  • Intent classifier 320 is configured to match a received utterance (e.g., utterance 306 or 308) to an intent associated with skill hot system 300.
  • a skill bot can be configured with one or more intents, each intent including at least one example utterance that is associated with the intent and used for training a classifier.
  • the intent classifier 242 of the master bot system 200 is trained to determine confidence scores for individual skill bots and confidence scores for system intents.
  • intent classifier 320 can be trained to determine a confidence score for each intent associated with the skill bot system 300.
  • the intent classifier 320 has access to intents information 354.
  • the intents information 354 includes, for each intent associated with the skill bot system 300, a list of utterances that are representative of and illustrate the meaning of the intent and are typically associated with a task performable by that intent.
  • the intents information 354 can further include parameters produced as a result of training on this list of utterances.
  • Conversation manager 330 receives, as an output of intent classifier 320, an indication 322 of a particular intent, identified by the intent classifier 320, as best matching the utterance that was input to the intent classifier 320.
  • the intent classifier 320 is unable to determine any match.
  • the confidence scores computed by the intent classifier 320 could fall below a threshold confidence score value if the utterance is directed to a system intent or an intent of a different skill hot.
  • the skill hot system 300 may refer the utterance to the master hot for handling, e.g., to route to a different skill hot.
  • the intent classifier 320 is successful in identifying an intent within the skill hot, then the conversation manager 330 will initiate a conversation with the user.
  • the conversation initiated by the conversation manager 330 is a conversation specific to the intent identified by the intent classifier 320.
  • the conversation manager 330 may be implemented using a state machine configured to execute a dialog flow for the identified intent.
  • the state machine can include a default starting state (e.g., for when the intent is invoked without any additional input) and one or more additional states, where each state has associated with it actions to be performed by the skill hot (e.g., executing a purchase transaction) and/or dialog (e.g., questions, responses) to be presented to the user.
  • the conversation manager 330 can determine an action/dialog 335 upon receiving the indication 322 identifying the intent and can determine additional actions or dialog in response to subsequent utterances received during the conversation.
  • Data store 350 comprises one or more computing devices that store data used by the various subsystems of the skill hot system 300. As depicted in FIG. 3, the data store 350 includes the rules 352 and the intents information 354. In certain aspects, data store 350 can be integrated into a data store of a master bot or digital assistant, e.g., the data store 250 in FIG. 2.
  • NER is a key NLP task that requires text analysis to understand relative semantics and sentiment. For example, knowing if a named entity like “Sydney” is a name of a place, a person, or a university is important to many natural language understanding tasks. Some of the important tasks that benefit from NER are described as follows.
  • NER can be used to identify the spans of text which constitutes answers. For example, the answer for the following question is “$95” based on the above example. How much did David Smith pay at the restaurant?
  • NER can identify entities as an initial step to know the customer’s sentiment towards the entities. For example, in the following sentence, there can be both positive and negative opinions about “Nikon camera” and “Canon camera” entities in the following example statement.
  • the Nikon camera is amazing; it is better than the Canon camera.
  • NER can be one of the technologies used in digital assistants. NER can be used in digital assistants to locate and classify the user’s words into predefined categories like PER (which can be short for PERSON), DATE and TIME. These types of information can be used by the digital assistant to process the user’s request. For example, to generate an expense report from the following sentence, a digital assistant can use an NER model to identify SFO, $10 and May 21st as MER (which can be short for MERCHANT), CUR (which can be short for CURRENCY) and DATE entities. LOC can be a category identifier short for LOCATION.
  • NER is a non-trivial task due to the difficulties caused by segmentation and type ambiguity. Segmentation ambiguity emerges from the complexity of finding the entities and their boundaries in the sentence. For example, multiple words in the sentence may represent one single entity. As demonstrated below, “New York Times” is a single entity (with MER tag) composed of 3 words:
  • a classifier can be a system which learns a function to determine the class or label among a predefined finite set of categories given a set of inputs.
  • sentiment analysis enables identifying the sentiment and orientation of an opinion expressed in a sentence (e.g., a review of a product) as being positive, negative, or neutral.
  • sequence labelling classifies (i. e. , assigns a label to) every word in a text; and NER as a sequence labelling task can include identifying and labelling subsequences of words.
  • Begins, Inside, and Outside (BIO) label scheme can be a way of using a sequence labeler to identify subsequences, where labels capture both the boundary and the type of named entity.
  • any token that begins a span of interest can be tagged with the label B
  • tokens that occur inside a span are tagged with the Label I
  • any tokens outside of any span of interest are labelled O. While there may be only one O tag, there can be different B and I tags for each named entity class (e.g., I-DATE and I-MER).
  • the following example shows how BIO encoding defines the boundary between two adjacent named entities of the same type (i.e., [MER Plaza Hotel’s] and [MER Palm Court Restaurant]). Table 1 illustrates an example categorization of an utterance.
  • An intelligent assistant NER can incorporate the following state-of-the-art technologies.
  • Pre-training refers to the process of training a network on large external datasets such as Wikipedia and Common Crawl (an open repository of freely provided data by crawling the web).
  • the intuition behind pre-training is that if a network is trained on a large and general enough dataset, it will be effectively served as a generic model of the textual world.
  • pre-trained networks we can exploit the knowledge gained while solving one problem (e.g., language modeling) to initialize the backbones of a new network for solving a different but related problem (e.g., text classification) without having to start from scratch (a.k.a., transfer learning).
  • Sequence labelling models can lead to inconsistency between neighboring labels.
  • the model has labelled “Sydney” as B-PER and “Harbour” as I-LOC which is not a valid sequence since the label boundary cannot start with an I tag. So, the correct sequence will be B-LOC and I-LOC. For this reason, conditional random field (CRF) is used to enforce that adjacent labels are consistent.
  • CRF conditional random field
  • Table 3 illustrates an exemplary labeling of a portion of a sentence.
  • CRF leams tag-tag weights while training and avoids generating impossible BIO tag sequences by assigning those sequences a very negative weight (e.g., X values in the following Table 3).
  • Deep learning models with a large number of parameters can easily be overfit to the training data. This has the effect of the model learning the noise in the training data which results in poor performance when the model is evaluated on new data. Dropout can be a computationally cheap and remarkably effective technique for addressing this problem. The key idea is to randomly ignore or drop out some layer output units from the model during training to prevent units from co-adapting too much (i.e., overfitting).
  • Deep learning models require different constraints and capacities to generalize well on different data patterns. These constraints are controlled by a number of measures called hyper-parameters which have to be tuned so that the model can optimally solve the problem. Hyper-parameter tuning refers to the process of finding a set of optimal hyper-parameters for a learning algorithm.
  • the disclosed NER system can include several model improvements, training improvements, and data improvements.
  • the disclosed NER system can combine the context and gazetteer features.
  • the disclosed NER model can be a hybrid model combining the context and gazetteer features.
  • the new method combines contextual features and external knowledge resources called gazetteers to improve the model performance.
  • Gazetteers can be lists of named entities such as organizations, countries, cities, and person names which are matched against unstructured text to provide additional features to the model. For example:
  • the disclosed NER system can include a fixed CRF tag-tag transitions on small training datasets.
  • CRF learns that inconsistent tag pairs like O I- LOC I-PER never appear in the training data, and hence assigns these tag-tag transitions a very negative weight.
  • CRF sometimes learns a model that generates these inconsistent tag-tag transitions. Therefore, the disclosed NER system can introduce a new technique to stop CRF from finding the inconsistent tag-tag transitions by modifying the corresponding weights after training by setting the weights for the inconsistent transitions to a very negative value. This technique guarantees generating consistent tag-tag transitions even with small training data.
  • the disclosed NER system can include selective dropout.
  • a new technique called selective dropout can apply a higher dropout rate to entity tokens compared to non-entity tokens. This can force the model to focus on the contextual information during training and makes the model more reliable and robust to different values for every entity type. For example: (The blue parts are entity values.)
  • the disclosed NER system can include extensive hyper-parameter tuning.
  • the hyper-tuning framework can cover a wide range of values. This enables the disclosed NER system to run extensive hyper-parameter tuning to identify an optimal choice of hyperparameters for achieving high-quality results.
  • the disclosed NER system can include data improvements.
  • One data improvement can include overlapping chunking mechanism.
  • a limitation of existing state-of-the-art models is that their memory and computational requirements grow quadratically with the input sequence length. Given the limitations of commonly available hardware, the current pretrained language models can only process input sequences of maximum 512 tokens.
  • the disclosed system can use a mechanism to break a long text into overlapping chunks, where each chunk and its corresponding labels are treated as a separate example when training. For evaluation/inference, we have to merge the predictions from chunks from the same input text. This mechanism enables the model to serve larger sequences and significantly reduces the training and inference time while maintaining high performance.
  • FIG. 4 illustrates a system 400 for handling long text for pre-trained language models.
  • the length determination engine 410 can receive the utterance 402.
  • the length determination engine can determine if the length exceeds a predetermined threshold of tokens.
  • Tokens can be a word, a portion of a word, or punctuation.
  • the chunking engine 420 can break the utterance into a plurality of overlapping chunks. For example, the chunking engine 420 can divide utterance 402 into Chunk-A 422, Chunk-B 424, and Chunk- C 426. While three chunks are illustrated, the disclosed techniques are not so limited and are applicable to any number of chunks. For example, as few as two chunks can be processed depending on the processing system capabilities.
  • the named entity recognizer 430 can determine a classifier or label for each of the chunks. For example, the named entity recognizer 430 can determine label-A 432 for Chunk- A 422, label-B 434 for Chunk-B 424, and label-C 436 for Chunk-C 426. As discussed above, named entity recognizer 430 can assign a label to each token in a chunk. The named entity recognizer can also assign a label to each chunk of the plurality of chunks. The score engine 440 can receive label-A 432 for Chunk-A 422, label-B 434 for Chunk-B 424, and label-C 436 for Chunk-C 426. Each of the labels can have an accompanying chunk score.
  • the Score Engine 440 can calculate a final label and an associated confidence score for each chunk of the overlapping chunks of token pieces by merging two confidence scores, one from a first chunk and another from a second chunk.
  • the Score Engine 440 can determine a final annotated label for the set of utterances based at least on the merging the two confidence scores.
  • the Score Engine 440 can store the final annotated score 442 in a memory.
  • FIG. 5 illustrates converting an exemplary utterance 502 into a plurality of chunks.
  • the example utterance 502 can be divided into a plurality of chunks, wherein each chunk is a certain number of words.
  • the example utterance 502 can be divided into segments of pre-determined lengths including a number of words. In some aspects, the predetermined length can be 32 words.
  • utterance 502 can be divided into Chunk-A 504 can include indices 1 through 10 for an exemplary chunk size of 10 tokens.
  • Chunk-B can include indices 6 through 15 for an exemplary chunk size of 10 tokens.
  • the named entity recognizer 430 as shown in FIG. 4, can process smaller chunks more efficiently. For example, it will be faster for the entity recognizer 430 to process the 10-word segments than it would be to process the entire example utterance 502.
  • the named entity recognizer 430 can analyze the utterance 502 and determine a tag or label for each of the tokens. For example, the first token of the utterance 502, at index 1, is the word “My” and the named entity recognizer 430 can assign the token the tag of “O” which means that it is outside of the range of interest. Similarly, the named entity recognizer 430 can assign the tokens, “name” “is” “,” “living” “in” “and” “working” “for” can be assigned a tag of “O”. The named entity recognizer 430 can assign the token “Davis” at index 4 a tag of “B-PER” meaning that is the beginning of a person’s name.
  • the named entity recognizer 430 can assign the token “Brisbane” at index 8 a tag of “B-LOC” meaning beginning of a location.
  • the named entity recognizer 430 can assign the token “Australia” at index 10 the tag of “B-LOC” meaning beginning of a location.
  • the named entity recognizer 430 can assign the token “Oracle” at index 14 the tag of “B-ORG” meaning the beginning of an organization.
  • the named entity recognizer 430 can assign the token “Corp” at index 15 the tag of inside the span of an organization name.
  • FIG. 5 illustrates a chunk size of 10 tokens with an overlapping size of 5 chunks.
  • a first overlap portion 508 can include the second half of Chunk-A 504.
  • a second overlap portion 510 can include the first half of Chunk-B 506.
  • FIG. 5 further illustrates techniques for merging predictions from the chunks.
  • the named entity recognizer 430 can properly label “Brisbane” at index 8 but misses “Australia” at index 10 during analysis of Chunk-A.
  • the named entity recognizer 430 can properly label “Australia” at index 10 but missed “Brisbane” at index 8.
  • the outputs of analysis from Chunk-A and Chunk-B are merged, but “Brisbane” and “Australia” and be properly labelled.
  • FIG. 6 illustrates techniques for merging the predictions.
  • the Score Engine 440 as shown in FIG. 4 can determine a confidence score for each of the labels from the named entity recognizer 430. For example, the Score Engine 440 can determine a confidence score for indices 6, 7, 8, and 9, the confidence score indicates that the confidence in the assigned labels being 0.9 representing a 90% confidence that the label is correct. For index 10 for Chunk-A, the Score Engine 440 can determine a confidence score for the assigned label of “O” being 0.5 representing a 50% confidence that the label is correct.
  • the Score Engine 440 can determine a confidence score for indices 6, 7, 9, and 10, the confidence score indicates that the confidence in the assigned labels being 0.9 representing a 90% confidence that the label is correct.
  • the Score Engine 440 can determine a confidence score for the assigned level of “O” being 0.5 representing a 50% confidence that the label is correct.
  • the output strategies can include “half,” “first,” “second,” and “max.”
  • the Score Engine 440 can use !4 prediction from the first chunk and !4 prediction from the second chunk.
  • the Score Engine 440 can use the prediction from the first chunk.
  • the Score Engine 440 can use the prediction from the second chunk.
  • the Score Engine 440 can decide on the prediction based on the maximum confidence scores.
  • the “half’ strategy can provide an accurate result because the labels for index 8 for “Brisbane” and label index 10 for “Australia” are both accurate.
  • the “first” strategy correctly identifies the label for index 8 for “Brisbane” but incorrectly labels the token piece for “Australia”.
  • the “second” strategy incorrectly identifies the label for index 8 for “Brisbane” but correctly labels the token piece for “Australia.”
  • the “Max” strategy can provide an accurate result because the labels for index 8 for “Brisbane” and label index 10 for “Australia” are both accurate.
  • the pre-determined length can be 32 token pieces. As shown in FIG. 6, each of the chunks can be broken up into multiple elements. For example, each of the chunks can provide overlap, which can be abbreviated as /, between the chunks. The amount of overlap between the chunks can be predetermined. In an example, for a 32 token piece chunk the overlap can be 16 token pieces (subwords).
  • FIG. 7 is a flowchart of an example process 700 associated with system and techniques for handling long text for pre-trained language models.
  • one or more process blocks of FIG. 7 may be performed by a computing device (e.g., computing device 1000).
  • one or more process blocks of FIG. 7 may be performed by another device, or a group of devices separate from or including the computing device. Additionally, or alternatively, one or more process blocks of FIG. 7 may be performed by one or more components of device 1000, such as processing subsystem 1004, storage subsystem 1018, I/O subsystem 1008, communication subsystem 1024, and/or bus subsystem 1002.
  • process 700 may include receiving, at a data processing system, a set of utterances for training or interfacing with a named entity recognizer to assign a label to each token piece from the set of utterances (block 710).
  • the computing device may receive, at a data processing system, a set of utterances for training or interfacing with a named entity recognizer to assign a label to each token piece from the set of utterances, as described above.
  • the set of utterances can be received through user entry via a Chabot using a keyboard.
  • the set of utterances can be received though user via a Chabot aurally using a microphone.
  • the system can convert the verbal utterance into a textual entry.
  • process 700 may include determining a length of the set of utterances (block 720).
  • the computing device may determine a length of the set of utterances, as described above.
  • the process 700 can determined the length of the set of utterances by dividing the utterance into a plurality of token pieces.
  • a token piece can be a word, a portion of a word, or punctuation.
  • Complex words can be broken down into one or more token pieces.
  • the length can be the number of token pieces in the utterance.
  • process 700 may include when the length of the set of utterances exceeds a pre-determined threshold of token pieces, the process 700 can include dividing the set of utterances into a plurality of overlapping chunks of token pieces (block 730).
  • the length of the utterance can be compared with a predetermined threshold.
  • the predetermined threshold can vary as required.
  • the pre-determined threshold can be 512 token pieces.
  • One skilled in the art would understand that other thresholds would be within the scope of the disclosure.
  • process 700 may include assigning a label together with a confidence score for each token piece in a chunk (block 740).
  • each of the token pieces in an utterance can be assigned a label.
  • the named entity recognizer 430 can identify and label subsequences of words.
  • a begins, inside, and outside (BIO) label scheme can used.
  • the BIO label scheme can identify subsequences, where labels capture both the boundary and the type of named entity.
  • any token that begins a span of interest can be tagged with the label B, tokens that occur inside a span are tagged with the label I, and any tokens outside of any span of interest are labelled O. While there may be only one O tag, there can be different B and I tags for each named entity class (e.g., I-DATE and I-MERCHANT).
  • Each label can be assigned a confidence score that indicates that confidence that the label is correct.
  • process 700 may include determining a final label and an associated confidence score for each chunk of the overlapping chunks of token pieces by merging two confidence scores, one from a first chunk and another from a second chunk (block 750).
  • process 700 may include determining a final annotated label for the set of utterances based at least on the merging the two confidence scores (block 760).
  • process 700 may include storing the final annotated label in a memory (block 770).
  • Process 700 can utilize several different strategies for determining a final annotated label.
  • the output strategies can include “half,” “first,” “second,” and “max.”
  • the Score Engine 440 can use !4 prediction from the first chunk and !4 prediction from the second chunk.
  • the Score Engine 440 can use the prediction from the first chunk.
  • the Score Engine 440 can use the prediction from the second chunk.
  • the Score Engine 440 can decide on the prediction based on the maximum confidence scores.
  • Process 700 may include additional implementations, such as any single implementation or any combination of implementations described below and/or in connection with one or more other processes described elsewhere herein.
  • each chunk and its corresponding sequence of labels are treated as a separate example when training.
  • the utterance with the length N is broken into [(N-L)/(K-L)] overlapping chunks wherein K is a chunk size and L is an overlapping size.
  • the determining an overall score of a token piece is based on a position of the token piece in the overlapping chunks of token pieces, wherein the overall score is a first confidence score from the first chunk if it is in a first half of in the overlapping chunks of token pieces and a second confidence score from the second chunk if it is in a second half of the overlapping chunks of token pieces.
  • the determining an overall score is based on a maximum confidence score of the plurality of overlapping chunks.
  • the pre-determined threshold of token pieces is 512 token pieces, and a pre-determined number of overlapping token pieces is 128 token pieces.
  • FIG. 7 shows example blocks of process 700
  • process 700 may include additional blocks, fewer blocks, different blocks, or differently arranged blocks than those depicted in FIG. 7. Additionally, or alternatively, two or more of the blocks of process 700 may be performed in parallel.
  • FIG. 8 depicts a simplified diagram of a distributed system 800 for implementing an embodiment.
  • distributed system 800 includes one or more client computing devices 802, 804, 806, and 808, coupled to a server 812 via one or more communication networks 810.
  • Clients computing devices 802, 804, 806, and 808 may be configured to execute one or more applications.
  • server 812 may be adapted to run one or more services or software applications that enable techniques for handling long text for pre-trained language models.
  • server 812 may also provide other services or software applications that can include non-virtual and virtual environments.
  • these services may be offered as web-based or cloud services, such as under a Software as a Service (SaaS) model to the users of client computing devices 802, 804, 806, and/or 808.
  • SaaS Software as a Service
  • Users operating client computing devices 802, 804, 806, and/or 808 may in turn utilize one or more client applications to interact with server 812 to utilize the services provided by these components.
  • Users may use client computing devices 802, 804, 806, and/or 808 for techniques for handling long text for pre-trained language models in accordance with the teachings of this disclosure.
  • a client device may provide an interface that enables a user of the client device to interact with the client device. The client device may also output information to the user via this interface.
  • FIG. 8 depicts only four client computing devices, any number of client computing devices may be supported.
  • the client devices may include various types of computing systems such as portable handheld devices, general purpose computers such as personal computers and laptops, workstation computers, wearable devices, gaming systems, thin clients, various messaging devices, sensors or other sensing devices, and the like. These computing devices may run various types and versions of software applications and operating systems (e.g., Microsoft Windows®, Apple Macintosh®, UNIX® or UNIX-like operating systems, Linux or Linux-like operating systems such as Google ChromeTM OS) including various mobile operating systems (e.g., Microsoft Windows Mobile®, iOS®, Windows Phone®, AndroidTM, BlackBerry®, Palm OS®).
  • Microsoft Windows Mobile® iOS®
  • Windows Phone® AndroidTM
  • BlackBerry® Palm OS®
  • Portable handheld devices may include cellular phones, smartphones, (e.g., an iPhone®), tablets (e.g., iPad®), personal digital assistants (PDAs), and the like.
  • Wearable devices may include Google Glass® head mounted display, and other devices.
  • Gaming systems may include various handheld gaming devices, Internet-enabled gaming devices (e.g., a Microsoft Xbox® gaming console with or without a Kinect® gesture input device, Sony PlayStation® system, various gaming systems provided by Nintendo®, and others), and the like.
  • the client devices may be capable of executing various different applications such as various Internet-related apps, communication applications (e.g., E-mail applications, short message service (SMS) applications) and may use various communication protocols.
  • communication applications e.g., E-mail applications, short message service (SMS) applications
  • Network(s) 810 may be any type of network familiar to those skilled in the art that can support data communications using any of a variety of available protocols, including without limitation TCP/IP (transmission control protocol/Intemet protocol), SNA (systems network architecture), IPX (Internet packet exchange), AppleTalk®, and the like.
  • TCP/IP transmission control protocol/Intemet protocol
  • SNA systems network architecture
  • IPX Internet packet exchange
  • AppleTalk® any type of network familiar to those skilled in the art that can support data communications using any of a variety of available protocols, including without limitation TCP/IP (transmission control protocol/Intemet protocol), SNA (systems network architecture), IPX (Internet packet exchange), AppleTalk®, and the like.
  • network(s) 810 can be a local area network (LAN), networks based on Ethernet, Token-Ring, a wide-area network (WAN), the Internet, a virtual network, a virtual private network (VPN), an intranet, an extranet, a public switched telephone network (PSTN), an infra-red network, a wireless network (e.g., a network operating under any of the Institute of Electrical and Electronics (IEEE) 1002.11 suite of protocols, Bluetooth®, and/or any other wireless protocol), and/or any combination of these and/or other networks.
  • LAN local area network
  • WAN wide-area network
  • VPN virtual private network
  • PSTN public switched telephone network
  • IEEE Institute of Electrical and Electronics
  • Server 812 may be composed of one or more general purpose computers, specialized server computers (including, by way of example, PC (personal computer) servers, UNIX® servers, mid-range servers, mainframe computers, rack-mounted servers, etc.), server farms, server clusters, or any other appropriate arrangement and/or combination.
  • Server 812 can include one or more virtual machines running virtual operating systems, or other computing architectures involving virtualization such as one or more flexible pools of logical storage devices that can be virtualized to maintain virtual storage devices for the server.
  • server 812 may be adapted to run one or more services or software applications that provide the functionality described in the foregoing disclosure.
  • server 812 may run one or more operating systems including any of those discussed above, as well as any commercially available server operating system.
  • Server 812 may also run any of a variety of additional server applications and/or mid-tier applications, including HTTP (hypertext transport protocol) servers, FTP (file transfer protocol) servers, CGI (common gateway interface) servers, JAVA® servers, database servers, and the like.
  • HTTP hypertext transport protocol
  • FTP file transfer protocol
  • CGI common gateway interface
  • JAVA® servers JAVA® servers
  • database servers and the like.
  • Exemplary database servers include without limitation those commercially available from Oracle®, Microsoft®, Sybase®, IBM® (International Business Machines), and the like.
  • server 812 may include one or more applications to analyze and consolidate data feeds and/or event updates received from users of client computing devices 802, 804, 806, and 808.
  • data feeds and/or event updates may include, but are not limited to, Twitter® feeds, Facebook® updates or real-time updates received from one or more third party information sources and continuous data streams, which may include real-time events related to sensor data applications, financial tickers, network performance measuring tools (e.g., network monitoring and traffic management applications), clickstream analysis tools, automobile traffic monitoring, and the like.
  • Server 812 may also include one or more applications to display the data feeds and/or real-time events via one or more display devices of client computing devices 802, 804, 806, and 808.
  • a data repository used by server 812 may be a database, for example, a relational database, such as databases provided by Oracle Corporation® and other vendors.
  • a relational database such as databases provided by Oracle Corporation® and other vendors.
  • One or more of these databases may be adapted to enable storage, update, and retrieval of data to and from the database in response to structured query language (SQL)-formatted commands.
  • SQL structured query language
  • one or more of data repositories 814, 816 may also be used by applications to store application data.
  • the data repositories used by applications may be of different types such as, for example, a key-value store repository, an object store repository, or a general storage repository supported by a file system.
  • FIG. 6 is a simplified block diagram of a cloud-based system environment in which various text handling-related services may be offered as cloud services, in accordance with certain aspects.
  • cloud infrastructure system 602 may provide one or more cloud services that may be requested by users using one or more client computing devices 604, 606, and 608.
  • Cloud infrastructure system 602 may comprise one or more computers and/or servers that may include those described above for server 812.
  • the computers in cloud infrastructure system 602 may be organized as general-purpose computers, specialized server computers, server farms, server clusters, or any other appropriate arrangement and/or combination.
  • Network(s) 610 may facilitate communication and exchange of data between clients 604, 606, and 608 and cloud infrastructure system 602.
  • Network(s) 610 may include one or more networks. The networks may be of the same or different types.
  • Network(s) 610 may support one or more communication protocols, including wired and/or wireless protocols, for facilitating the communications.
  • cloud infrastructure system 902 may have more or fewer components than those depicted in FIG. 9, may combine two or more components, or may have a different configuration or arrangement of components.
  • FIG. 9 depicts three client computing devices, any number of client computing devices may be supported in alternative aspects.
  • cloud service is generally used to refer to a service that is made available to users on demand and via a communication network such as the Internet by systems (e.g., cloud infrastructure system 902) of a service provider.
  • systems e.g., cloud infrastructure system 902
  • cloud service provider typically, in a public cloud environment, servers and systems that make up the cloud service provider's system are different from the customer's own on-premise servers and systems.
  • the cloud service provider’s systems are managed by the cloud service provider. Customers can thus avail themselves of cloud services provided by a cloud service provider without having to purchase separate licenses, support, or hardware and software resources for the services.
  • a cloud service provider's system may host an application, and a user may, via a network 910 (e.g., the Internet), on demand, order and use the application without the user having to buy infrastructure resources for executing the application.
  • Cloud services are designed to provide easy, scalable access to applications, resources and services.
  • Several providers offer cloud services. For example, several cloud services are offered by Oracle Corporation® of Redwood Shores, California, such as middleware services, database services, Java cloud services, and others.
  • cloud infrastructure system 902 may provide one or more cloud services using different models such as under a Software as a Service (SaaS) model, a Platform as a Service (PaaS) model, an Infrastructure as a Service (laaS) model, and others, including hybrid service models.
  • SaaS Software as a Service
  • PaaS Platform as a Service
  • laaS Infrastructure as a Service
  • Cloud infrastructure system 902 may include a suite of applications, middleware, databases, and other resources that enable provision of the various cloud services.
  • a SaaS model enables an application or software to be delivered to a customer over a communication network like the Internet, as a service, without the customer having to buy the hardware or software for the underlying application.
  • a SaaS model may be used to provide customers access to on-demand applications that are hosted by cloud infrastructure system 902.
  • Examples of SaaS services provided by Oracle Corporation® include, without limitation, various services for human resources/capital management, customer relationship management (CRM), enterprise resource planning (ERP), supply chain management (SCM), enterprise performance management (EPM), analytics services, social applications, and others.
  • An laaS model is generally used to provide infrastructure resources (e.g., servers, storage, hardware and networking resources) to a customer as a cloud service to provide elastic compute and storage capabilities.
  • infrastructure resources e.g., servers, storage, hardware and networking resources
  • Various laaS services are provided by Oracle Corporation®.
  • a PaaS model is generally used to provide, as a service, platform and environment resources that enable customers to develop, run, and manage applications and services without the customer having to procure, build, or maintain such resources.
  • PaaS services provided by Oracle Corporation® include, without limitation, Oracle Java Cloud Service (JCS), Oracle Database Cloud Service (DBCS), data management cloud service, various application development solutions services, and others.
  • Cloud services are generally provided on an on-demand self-service basis, subscription-based, elastically scalable, reliable, highly available, and secure manner.
  • a customer via a subscription order, may order one or more services provided by cloud infrastructure system 902.
  • Cloud infrastructure system 902 then performs processing to provide the services requested in the customer's subscription order.
  • Cloud infrastructure system 902 may be configured to provide one or even multiple cloud services.
  • Cloud infrastructure system 902 may provide the cloud services via different deployment models.
  • cloud infrastructure system 902 may be owned by a third-party cloud services provider and the cloud services are offered to any general public customer, where the customer can be an individual or an enterprise.
  • cloud infrastructure system 902 may be operated within an organization (e.g., within an enterprise organization) and services provided to customers that are within the organization.
  • the customers may be various departments of an enterprise such as the Human Resources department, the Payroll department, etc. or even individuals within the enterprise.
  • the cloud infrastructure system 902 and the services provided may be shared by several organizations in a related community.
  • Various other models such as hybrids of the above- mentioned models may also be used.
  • Client computing devices 904, 906, and 908 may be of different types (such as devices 802, 804, 806, and 808 depicted in FIG. 8) and may be capable of operating one or more client applications.
  • a user may use a client device to interact with cloud infrastructure system 902, such as to request a service provided by cloud infrastructure system 902.
  • cloud infrastructure system 902. For example, a user may use a client device to request a chat hot service described in this disclosure.
  • the processing performed by cloud infrastructure system 902 for providing Chabot services may involve big data analysis.
  • This analysis may involve using, analyzing, and manipulating large data sets to detect and visualize various trends, behaviors, relationships, etc. within the data.
  • This analysis may be performed by one or more processors, possibly processing the data in parallel, performing simulations using the data, and the like.
  • big data analysis may be performed by cloud infrastructure system 902 for determining the intent of an utterance.
  • the data used for this analysis may include structured data (e.g., data stored in a database or structured according to a structured model) and/or unstructured data (e.g., data blobs (binary large objects)).
  • cloud infrastructure system 902 may include infrastructure resources 930 that are utilized for facilitating the provision of various cloud services offered by cloud infrastructure system 902.
  • Infrastructure resources 930 may include, for example, processing resources, storage or memory resources, networking resources, and the like.
  • the resources may be bundled into sets of resources or resource modules (also referred to as "pods").
  • Each resource module or pod may comprise a pre-integrated and optimized combination of resources of one or more types.
  • different pods may be pre-provisioned for different types of cloud services. For example, a first set of pods may be provisioned for a database service, a second set of pods, which may include a different combination of resources than a pod in the first set of pods, may be provisioned for Java service, and the like.
  • the resources allocated for provisioning the services may be shared between the services.
  • Cloud infrastructure system 902 may itself internally use services 932 that are shared by different components of cloud infrastructure system 902 and which facilitate the provisioning of services by cloud infrastructure system 902.
  • These internal shared services may include, without limitation, a security and identity service, an integration service, an enterprise repository service, an enterprise manager service, a virus scanning and whitelist service, a high availability, backup and recovery service, service for enabling cloud support, an email service, a notification service, a file transfer service, and the like.
  • Cloud infrastructure system 902 may comprise multiple subsystems. These subsystems may be implemented in software, or hardware, or combinations thereof. As depicted in FIG. 9, the subsystems may include a user interface subsystem 912 that enables users or customers of cloud infrastructure system 902 to interact with cloud infrastructure system 902. User interface subsystem 912 may include various different interfaces such as a web interface 914, an online store interface 916 where cloud services provided by cloud infrastructure system 902 are advertised and are purchasable by a consumer, and other interfaces 918. For example, a customer may, using a client device, request (service request 934) one or more services provided by cloud infrastructure system 902 using one or more of interfaces 914, 916, and 918.
  • request service request 934
  • a customer may access the online store, browse cloud services offered by cloud infrastructure system 902, and place a subscription order for one or more services offered by cloud infrastructure system 902 that the customer wishes to subscribe to.
  • the service request may include information identifying the customer and one or more services that the customer desires to subscribe to.
  • a customer may place a subscription order for a Chabot related service offered by cloud infrastructure system 902.
  • the customer may provide information identifying for input (e.g., utterances).
  • cloud infrastructure system 902 may comprise an order management subsystem (OMS) 920 that is configured to process the new order.
  • OMS 920 may be configured to: create an account for the customer, if not done already; receive billing and/or accounting information from the customer that is to be used for billing the customer for providing the requested service to the customer; verify the customer information; upon verification, book the order for the customer; and orchestrate various workflows to prepare the order for provisioning.
  • OMS 920 may then invoke the order provisioning subsystem (OPS) 924 that is configured to provision resources for the order including processing, memory, and networking resources.
  • the provisioning may include allocating resources for the order and configuring the resources to facilitate the service requested by the customer order.
  • the manner in which resources are provisioned for an order and the type of the provisioned resources may depend upon the type of cloud service that has been ordered by the customer.
  • OPS 924 may be configured to determine the particular cloud service being requested and identify a number of pods that may have been pre-configured for that particular cloud service. The number of pods that are allocated for an order may depend upon the size/amount/level/scope of the requested service.
  • the number of pods to be allocated may be determined based upon the number of users to be supported by the service, the duration of time for which the service is being requested, and the like. The allocated pods may then be customized for the particular requesting customer for providing the requested service.
  • Cloud infrastructure system 902 may send a response or notification 944 to the requesting customer to indicate when the requested service is now ready for use. In some instances, information (e.g., a link) may be sent to the customer that enables the customer to start using and availing the benefits of the requested services.
  • Cloud infrastructure system 902 may provide services to multiple customers. For each customer, cloud infrastructure system 902 is responsible for managing information related to one or more subscription orders received from the customer, maintaining customer data related to the orders, and providing the requested services to the customer. Cloud infrastructure system 902 may also collect usage statistics regarding a customer's use of subscribed services. For example, statistics may be collected for the amount of storage used, the amount of data transferred, the number of users, and the amount of system up time and system down time, and the like. This usage information may be used to bill the customer. Billing may be done, for example, on a monthly cycle.
  • Cloud infrastructure system 902 may provide services to multiple customers in parallel. Cloud infrastructure system 902 may store information for these customers, including possibly proprietary information.
  • cloud infrastructure system 902 comprises an identity management subsystem (IMS) 928 that is configured to manage customers information and provide the separation of the managed information such that information related to one customer is not accessible by another customer.
  • IMS 928 may be configured to provide various security -related services such as identity services, such as information access management, authentication and authorization services, services for managing customer identities and roles and related capabilities, and the like.
  • FIG. 10 illustrates an exemplary computer system 1000 that may be used to implement certain aspects.
  • computer system 1000 may be used to implement any of the system 400 for handling long text for pre-trained language models as shown in FIG 4. and various servers and computer systems described above.
  • computer system 1000 includes various subsystems including a processing subsystem 1004 that communicates with a number of other subsystems via a bus subsystem 1002. These other subsystems may include a processing acceleration unit 1006, an I/O subsystem 1008, a storage subsystem 1018, and a communications subsystem 1024.
  • Storage subsystem 1018 may include non-transitory computer-readable storage media including storage media 1022 and a system memory 1010.
  • Bus subsystem 1002 provides a mechanism for letting the various components and subsystems of computer system 1000 communicate with each other as intended. Although bus subsystem 1002 is shown schematically as a single bus, alternative aspects of the bus subsystem may utilize multiple buses. Bus subsystem 1002 may be any of several types of bus structures including a memory bus or memory controller, a peripheral bus, a local bus using any of a variety of bus architectures, and the like.
  • such architectures may include an Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MCA) bus, Enhanced ISA (EISA) bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus, which can be implemented as a Mezzanine bus manufactured to the IEEE P1386.1 standard, and the like.
  • ISA Industry Standard Architecture
  • MCA Micro Channel Architecture
  • EISA Enhanced ISA
  • VESA Video Electronics Standards Association
  • PCI Peripheral Component Interconnect
  • Processing subsystem 1004 controls the operation of computer system 1000 and may comprise one or more processors, application specific integrated circuits (ASICs), or field programmable gate arrays (FPGAs).
  • the processors may include be single core or multi core processors.
  • the processing resources of computer system 1000 can be organized into one or more processing units 1032, 1034, etc.
  • a processing unit may include one or more processors, one or more cores from the same or different processors, a combination of cores and processors, or other combinations of cores and processors.
  • processing subsystem 1004 can include one or more special purpose co-processors such as graphics processors, digital signal processors (DSPs), or the like.
  • DSPs digital signal processors
  • some or all of the processing units of processing subsystem 1004 can be implemented using customized circuits, such as application specific integrated circuits (ASICs), or field programmable gate arrays (FPGAs).
  • processing units in processing subsystem 1004 can execute instructions stored in system memory 1010 or on computer readable storage media 1022.
  • the processing units can execute a variety of programs or code instructions and can maintain multiple concurrently executing programs or processes. At any given time, some or all of the program code to be executed can be resident in system memory 1010 and/or on computer-readable storage media 1022 including potentially on one or more storage devices.
  • processing subsystem 1004 can provide various functionalities described above. In instances where computer system 1000 is executing one or more virtual machines, one or more processing units may be allocated to each virtual machine.
  • a processing acceleration unit 1006 may optionally be provided for performing customized processing or for off-loading some of the processing performed by processing subsystem 1004 so as to accelerate the overall processing performed by computer system 1000.
  • I/O subsystem 1008 may include devices and mechanisms for inputting information to computer system 1000 and/or for outputting information from or via computer system 1000.
  • input device is intended to include all possible types of devices and mechanisms for inputting information to computer system 1000.
  • User interface input devices may include, for example, a keyboard, pointing devices such as a mouse or trackball, a touchpad or touch screen incorporated into a display, a scroll wheel, a click wheel, a dial, a button, a switch, a keypad, audio input devices with voice command recognition systems, microphones, and other types of input devices.
  • User interface input devices may also include motion sensing and/or gesture recognition devices such as the Microsoft Kinect® motion sensor that enables users to control and interact with an input device, the Microsoft Xbox® 360 game controller, devices that provide an interface for receiving input using gestures and spoken commands.
  • User interface input devices may also include eye gesture recognition devices such as the Google Glass® blink detector that detects eye activity (e.g., "blinking" while taking pictures and/or making a menu selection) from users and transforms the eye gestures as inputs to an input device (e.g., Google Glass®).
  • user interface input devices may include voice recognition sensing devices that enable users to interact with voice recognition systems (e.g., Siri® navigator) through voice commands.
  • user interface input devices include, without limitation, three dimensional (3D) mice, joysticks or pointing sticks, gamepads and graphic tablets, and audio/visual devices such as speakers, digital cameras, digital camcorders, portable media players, webcams, image scanners, fingerprint scanners, barcode reader 3D scanners, 3D printers, laser rangefinders, and eye gaze tracking devices.
  • user interface input devices may include, for example, medical imaging input devices such as computed tomography, magnetic resonance imaging, position emission tomography, and medical ultrasonography devices.
  • User interface input devices may also include, for example, audio input devices such as MIDI keyboards, digital musical instruments and the like.
  • output device is intended to include all possible types of devices and mechanisms for outputting information from computer system 1000 to a user or other computer.
  • User interface output devices may include a display subsystem, indicator lights, or non-visual displays such as audio output devices, etc.
  • the display subsystem may be a cathode ray tube (CRT), a flat-panel device, such as that using a liquid crystal display (LCD) or plasma display, a projection device, a touch screen, and the like.
  • CTR cathode ray tube
  • LCD liquid crystal display
  • plasma display a projection device
  • touch screen a touch screen
  • user interface output devices may include, without limitation, a variety of display devices that visually convey text, graphics and audio/video information such as monitors, printers, speakers, headphones, automotive navigation systems, plotters, voice output devices, and modems.
  • Storage subsystem 1018 provides a repository or data store for storing information and data that is used by computer system 1000.
  • Storage subsystem 1018 provides a tangible non-transitory computer-readable storage medium for storing the basic programming and data constructs that provide the functionality of some aspects.
  • Storage subsystem 1018 may store software (e.g., programs, code modules, instructions) that when executed by processing subsystem 1004 provides the functionality described above.
  • the software may be executed by one or more processing units of processing subsystem 1004.
  • Storage subsystem 1018 may also provide a repository for storing data used in accordance with the teachings of this disclosure.
  • Storage subsystem 1018 may include one or more non-transitory memory devices, including volatile and non-volatile memory devices. As shown in FIG. 10, storage subsystem 1018 includes a system memory 1010 and a computer-readable storage media 1022.
  • System memory 1010 may include a number of memories including a volatile main random-access memory (RAM) for storage of instructions and data during program execution and a nonvolatile read only memory (ROM) or flash memory in which fixed instructions are stored.
  • RAM main random-access memory
  • ROM read only memory
  • BIOS basic input/output system
  • the RAM typically contains data and/or program modules that are presently being operated and executed by processing subsystem 1004.
  • system memory 1010 may include multiple different types of memory, such as static random-access memory (SRAM), dynamic random access memory (DRAM), and the like.
  • SRAM static random-access memory
  • DRAM dynamic random access memory
  • system memory 1010 may load application programs 1012 that are being executed, which may include various applications such as Web browsers, mid-tier applications, relational database management systems (RDBMS), etc., program data 1014, and an operating system 1016.
  • application programs 1012 may include various applications such as Web browsers, mid-tier applications, relational database management systems (RDBMS), etc.
  • program data 1014 and an operating system 1016.
  • operating system 1016 may include various versions of Microsoft Windows®, Apple Macintosh®, and/or Linux operating systems, a variety of commercially - available UNIX® or UNIX-like operating systems (including without limitation the variety of GNU/Linux operating systems, the Google Chrome® OS, and the like) and/or mobile operating systems such as iOS, Windows® Phone, Android® OS, BlackBerry® OS, Palm® OS operating systems, and others.
  • Computer-readable storage media 1022 may store programming and data constructs that provide the functionality of some aspects.
  • Computer-readable media 1022 may provide storage of computer-readable instructions, data structures, program modules, and other data for computer system 1000.
  • Software programs, code modules, instructions that, when executed by processing subsystem 1004 provides the functionality described above, may be stored in storage subsystem 1018.
  • computer-readable storage media 1022 may include non-volatile memory such as a hard disk drive, a magnetic disk drive, an optical disk drive such as a CD ROM, digital video disc (DVD), a Blu-Ray® disk, or other optical media.
  • Computer-readable storage media 1022 may include, but is not limited to, Zip® drives, flash memory cards, universal serial bus (USB) flash drives, secure digital (SD) cards, DVD disks, digital video tape, and the like.
  • Computer-readable storage media 1022 may also include, solid-state drives (SSD) based on non-volatile memory such as flashmemory based SSDs, enterprise flash drives, solid state ROM, and the like, SSDs based on volatile memory such as solid state RAM, dynamic RAM, static RAM, dynamic random access memory (DRAM)-based SSDs, magnetoresistive RAM (MRAM) SSDs, and hybrid SSDs that use a combination of DRAM and flash memory based SSDs.
  • SSD solid-state drives
  • DRAM dynamic random access memory
  • MRAM magnetoresistive RAM
  • storage subsystem 1018 may also include a computer-readable storage media reader 1020 that can further be connected to computer-readable storage media 1022.
  • Reader 1020 may receive and be configured to read data from a memory device such as a disk, a flash drive, etc.
  • computer system 1000 may support virtualization technologies, including but not limited to virtualization of processing and memory resources.
  • computer system 1000 may provide support for executing one or more virtual machines.
  • computer system 1000 may execute a program such as a hypervisor that facilitated the configuring and managing of the virtual machines.
  • Each virtual machine may be allocated memory, compute (e.g., processors, cores), I/O, and networking resources.
  • Each virtual machine generally runs independently of the other virtual machines.
  • a virtual machine typically runs its own operating system, which may be the same as or different from the operating systems executed by other virtual machines executed by computer system 1000. Accordingly, multiple operating systems may potentially be run concurrently by computer system 1000.
  • Communications subsystem 1024 provides an interface to other computer systems and networks. Communications subsystem 1024 serves as an interface for receiving data from and transmitting data to other systems from computer system 1000. For example, communications subsystem 1024 may enable computer system 1000 to establish a communication channel to one or more client devices via the Internet for receiving and sending information from and to the client devices. For example, the communication subsystem may be used to transmit a response to a user regarding the inquiry for a Chabot.
  • Communication subsystem 1024 may support both wired and/or wireless communication protocols.
  • communications subsystem 1024 may include radio frequency (RF) transceiver components for accessing wireless voice and/or data networks (e.g., using cellular telephone technology, advanced data network technology, such as 3G, 4G or EDGE (enhanced data rates for global evolution), Wi-Fi (IEEE 802.XX family standards, or other mobile communication technologies, or any combination thereol), global positioning system (GPS) receiver components, and/or other components.
  • RF radio frequency
  • communications subsystem 1024 can provide wired network connectivity (e.g., Ethernet) in addition to or instead of a wireless interface.
  • Communication subsystem 1024 can receive and transmit data in various forms.
  • communications subsystem 1024 may receive input communications in the form of structured and/or unstructured data feeds 1026, event streams 1028, event updates 1030, and the like.
  • communications subsystem 1024 may be configured to receive (or send) data feeds 1026 in real-time from users of social media networks and/or other communication services such as Twitter® feeds, Facebook® updates, web feeds such as Rich Site Summary (RSS) feeds, and/or real-time updates from one or more third party information sources.
  • RSS Rich Site Summary
  • communications subsystem 1024 may be configured to receive data in the form of continuous data streams, which may include event streams 1028 of realtime events and/or event updates 1030, that may be continuous or unbounded in nature with no explicit end.
  • applications that generate continuous data may include, for example, sensor data applications, financial tickers, network performance measuring tools (e.g., network monitoring and traffic management applications), clickstream analysis tools, automobile traffic monitoring, and the like.
  • Communications subsystem 1024 may also be configured to communicate data from computer system 1000 to other computer systems or networks.
  • the data may be communicated in various different forms such as structured and/or unstructured data feeds 1026, event streams 1028, event updates 1030, and the like to one or more databases that may be in communication with one or more streaming data source computers coupled to computer system 1000.
  • Computer system 1000 can be one of various types, including a handheld portable device (e.g., an iPhone® cellular phone, an iPad® computing tablet, a personal digital assistant (PDA)), a wearable device (e.g., a Google Glass® head mounted display), a personal computer, a workstation, a mainframe, a kiosk, a server rack, or any other data processing system.
  • a handheld portable device e.g., an iPhone® cellular phone, an iPad® computing tablet, a personal digital assistant (PDA)
  • PDA personal digital assistant
  • a wearable device e.g., a Google Glass® head mounted display
  • a personal computer e.g., a workstation, a mainframe, a kiosk, a server rack, or any other data processing system.
  • FIG. 10 Due to the ever-changing nature of computers and networks, the description of computer system 1000 depicted in FIG. 10 is intended only as a specific example. Many other configurations having more or fewer components than the system depicted in FIG
  • Such configuration can be accomplished, for example, by designing electronic circuits to perform the operation, by programming programmable electronic circuits (such as microprocessors) to perform the operation such as by executing computer instructions or code, or processors or cores programmed to execute code or instructions stored on a non-transitory memory medium, or any combination thereof.
  • Processes can communicate using a variety of techniques including but not limited to conventional techniques for inter-process communications, and different pairs of processes may use different techniques, or the same pair of processes may use different techniques at different times.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)

Abstract

Selon certains aspects, un dispositif informatique peut recevoir, au niveau d'un système de traitement de données, un ensemble d'énoncés à des fins d'entraînement ou d'inférence avec un dispositif de reconnaissance d'entité nommée pour attribuer une étiquette à chaque élément de jeton à partir de l'ensemble d'énoncés. Le dispositif informatique peut déterminer une longueur de chaque énoncé dans l'ensemble et, lorsque la longueur de l'énoncé dépasse un seuil prédéterminé d'éléments de jeton : diviser l'énoncé en une pluralité de blocs chevauchants d'éléments de jeton ; attribuer une étiquette avec un score de confiance pour chaque élément de jeton dans un bloc ; déterminer une étiquette finale et un score de confiance associé pour chaque bloc d'éléments de jeton par fusion de deux scores de confiance ; déterminer une étiquette annotée finale pour l'énoncé sur la base au moins de la fusion des deux scores de confiance ; et stocker l'étiquette annotée finale dans une mémoire.
PCT/US2022/050024 2021-11-22 2022-11-16 Système et techniques de manipulation d'un texte long pour des modèles de langage préentraînés WO2023091436A1 (fr)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US202163282146P 2021-11-22 2021-11-22
US63/282,146 2021-11-22
US17/750,240 2022-05-20
US17/750,240 US20230161963A1 (en) 2021-11-22 2022-05-20 System and techniques for handling long text for pre-trained language models

Publications (1)

Publication Number Publication Date
WO2023091436A1 true WO2023091436A1 (fr) 2023-05-25

Family

ID=84537666

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2022/050024 WO2023091436A1 (fr) 2021-11-22 2022-11-16 Système et techniques de manipulation d'un texte long pour des modèles de langage préentraînés

Country Status (1)

Country Link
WO (1) WO2023091436A1 (fr)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021202552A1 (fr) * 2020-03-30 2021-10-07 Oracle International Corporation Techniques améliorées pour la détection hors domaine (ood)

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021202552A1 (fr) * 2020-03-30 2021-10-07 Oracle International Corporation Techniques améliorées pour la détection hors domaine (ood)

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
BEVAN ROBERT ET AL: "Fine-tuning transformers for automatic chemical entity identification in PubMed articles", PROCEEDINGS OF THE BIOCREATIVE VII CHALLENGE EVALUATION WORKSHOP, 8 November 2021 (2021-11-08) - 2 February 2023 (2023-02-02), XP093020227, Retrieved from the Internet <URL:https://biocreative.bioinformatics.udel.edu/media/store/files/2021/TRACK2_pos_09_BC7_submission_135.pdf> [retrieved on 20230202] *
JOSHI MANDAR ET AL: "BERT for Coreference Resolution: Baselines and Analysis", PROCEEDINGS OF THE 2019 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING AND THE 9TH INTERNATIONAL JOINT CONFERENCE ON NATURAL LANGUAGE PROCESSING (EMNLP-IJCNLP), 22 December 2019 (2019-12-22), Stroudsburg, PA, USA, pages 5802 - 5807, XP055918047, Retrieved from the Internet <URL:https://arxiv.org/pdf/1908.09091.pdf> [retrieved on 20230202], DOI: 10.18653/v1/D19-1588 *
RUI ZHANG ET AL: "NameRec*: Highly Accurate and Fine-grained Person Name Recognition", ARXIV.ORG, 23 March 2021 (2021-03-23), XP081913944, Retrieved from the Internet <URL:https://arxiv.org/pdf/2103.11360.pdf> [retrieved on 20230202] *
THOMAS VAN DONGEN ET AL: "SChuBERT: Scholarly Document Chunks with BERT-encoding boost Citation Count Prediction", ARXIV.ORG, CORNELL UNIVERSITY LIBRARY, 201 OLIN LIBRARY CORNELL UNIVERSITY ITHACA, NY 14853, 22 December 2020 (2020-12-22), XP081843941, DOI: 10.18653/V1/2020.SDP-1.17 *

Similar Documents

Publication Publication Date Title
US11651768B2 (en) Stop word data augmentation for natural language processing
US12014146B2 (en) Techniques for out-of-domain (OOD) detection
US11868727B2 (en) Context tag integration with named entity recognition models
US20220230000A1 (en) Multi-factor modelling for natural language processing
EP4128010A1 (fr) Augmentation de données de bruit pour traitement du langage naturel
US20230100508A1 (en) Fusion of word embeddings and word scores for text classification
US20230205999A1 (en) Gazetteer integration for neural named entity recognition
US20230095673A1 (en) Extracting key information from document using trained machine-learning models
WO2022115727A1 (fr) Logits améliorés pour le traitement automatique du langage naturel
US20230098783A1 (en) Framework for focused training of language models and techniques for end-to-end hypertuning of the framework
US20230139397A1 (en) Deep learning techniques for extraction of embedded data from documents
US20230115321A1 (en) Fine-tuning multi-head network from a single transformer layer of pre-trained language model
US20220229991A1 (en) Multi-feature balancing for natural language processors
WO2022115736A1 (fr) Valeurs de logit basées sur une distance pour le traitement automatique du langage naturel
US20230161963A1 (en) System and techniques for handling long text for pre-trained language models
US20230136965A1 (en) Prohibiting inconsistent named entity recognition tag sequences
US12026468B2 (en) Out-of-domain data augmentation for natural language processing
US20230154455A1 (en) Path dropout for natural language processing
US20230206125A1 (en) Lexical dropout for natural language processing
WO2023091436A1 (fr) Système et techniques de manipulation d&#39;un texte long pour des modèles de langage préentraînés
US20240062112A1 (en) Adaptive training data augmentation to facilitate training named entity recognition models
US20230419052A1 (en) Techniques for positive entity aware augmentation using two-stage augmentation
US20220171938A1 (en) Out-of-domain data augmentation for natural language processing
US20230134149A1 (en) Rule-based techniques for extraction of question and answer pairs from data
WO2024044491A1 (fr) Augmentation adaptative de données d&#39;apprentissage pour faciliter l&#39;apprentissage de modèles de reconnaissance d&#39;entité nommée

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22826518

Country of ref document: EP

Kind code of ref document: A1