CN112926295A - Model recommendation method and device - Google Patents
Model recommendation method and device Download PDFInfo
- Publication number
- CN112926295A CN112926295A CN202110338638.6A CN202110338638A CN112926295A CN 112926295 A CN112926295 A CN 112926295A CN 202110338638 A CN202110338638 A CN 202110338638A CN 112926295 A CN112926295 A CN 112926295A
- Authority
- CN
- China
- Prior art keywords
- model
- text
- determining
- information
- demand
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
- G06F40/186—Templates
-
- 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
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
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)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a model recommendation method and device, and relates to the technical field of artificial intelligence. One embodiment of the method comprises: determining model requirement information, the model requirement information comprising at least one of: model name, model classification and model text information; determining a demand text vector corresponding to the model demand information; for each candidate model: determining a model text vector corresponding to the candidate model; determining a similarity value between the demand text vector and the model text vector; and determining a recommendation result corresponding to the model demand information according to the similarity value. The implementation method does not need to carry out model recommendation according to the historical behaviors of the user, and has a good model recommendation effect.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a model recommendation method and device.
Background
And the model recommendation is used for quickly finding out the relevant models preferred by the users from the massive model library according to the requirements of the users, so that the users can quickly reuse and reference the relevant models to shorten the research and development period. The traditional model recommendation method is usually required to recommend according to the historical behaviors of the user, and the cold start problem cannot be well solved. For example, when a new user has not yet acted in the system, the user cannot be recommended the model he wants.
Disclosure of Invention
In view of this, embodiments of the present invention provide a model recommendation method and apparatus, which do not need to perform model recommendation according to a historical behavior of a user and have a better model recommendation effect.
In a first aspect, an embodiment of the present invention provides a model recommendation method, including:
determining model requirement information, the model requirement information comprising at least one of: model name, model classification and model text information;
determining a demand text vector corresponding to the model demand information;
for each candidate model: determining a model text vector corresponding to the candidate model; determining a similarity value between the demand text vector and the model text vector;
and determining a recommendation result corresponding to the model demand information according to the similarity value.
Optionally, the determining the model requirement information includes:
receiving a model recommendation request;
and determining the model demand information according to the model recommendation request.
Optionally, the determining a requirement text vector corresponding to the model requirement information includes:
generating a model demand text corresponding to the model demand information;
and determining a demand text vector corresponding to the model demand information according to the model demand text.
Optionally, the determining, according to the model requirement text, a requirement text vector corresponding to the model requirement information includes:
preprocessing the model requirement text to generate a preprocessed text;
determining the weight of each term in the preprocessed text;
and generating a demand text vector corresponding to the model demand information according to the weight of the lexical item.
Optionally, the preprocessing the model requirement text to generate a preprocessed text includes:
performing word segmentation processing on the model requirement text;
carrying out stop word filtering processing on the model requirement text after word segmentation processing;
and carrying out synonym conversion processing on the filtered model requirement text to generate the preprocessed text.
Optionally, the determining the weight of each term in the preprocessed text includes:
determining a plurality of target terms from the preprocessed text;
and respectively determining the weight of each target term by utilizing a tf-dif algorithm.
Optionally, the generating a requirement text vector corresponding to the model requirement information according to the weight of the term includes:
determining a feature word set of the predicted text according to the weight of the lexical item, wherein the feature word set comprises at least one feature lexical item;
acquiring a preset keyword set;
and generating a demand text vector corresponding to the model demand information according to the feature word set and the keyword set.
Optionally, the determining, according to the similarity value, a recommendation result corresponding to the model requirement information includes:
selecting a preset number of candidate models from the candidate models according to the similarity value;
and generating a recommendation result corresponding to the model demand information according to the preset number of candidate models.
Optionally, before determining the model text vector corresponding to the candidate model, the method further includes:
and generating a model text vector corresponding to the candidate model.
Optionally, the generating a model text vector corresponding to the candidate model includes:
obtaining model description information of the candidate model;
generating a model description text corresponding to the model description information;
and determining a model text vector corresponding to the candidate model according to the model description text.
Optionally, the model description information includes: model attribute information and model package information;
the generating of the model description text corresponding to the model description information includes:
extracting a plurality of terms from the model packet information;
determining weights of the plurality of terms respectively;
determining at least one keyword term of the model package according to the weight of the term;
and generating a model description text corresponding to the model description information according to the at least one keyword and the model attribute information.
Optionally, the model attribute information includes at least one of: model name, model classification, model output parameters, model input parameters, application objects and evaluation indexes.
In a second aspect, an embodiment of the present invention provides a model recommendation apparatus, including:
an information determination module configured to determine model requirement information, the model requirement information including at least one of: model name, model classification and model text information;
the vector determination module is used for determining a demand text vector corresponding to the model demand information;
a similarity value determination module for, for each candidate model: determining a model text vector corresponding to the candidate model; determining a similarity value between the demand text vector and the model text vector;
and the result determining module is used for determining a recommendation result corresponding to the model demand information according to the similarity value.
In a third aspect, an embodiment of the present invention provides an electronic device, including:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the method of any of the embodiments described above.
In a fourth aspect, an embodiment of the present invention provides a computer-readable medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the method of any one of the above embodiments.
One embodiment of the above invention has the following advantages or benefits: and determining a recommendation result corresponding to the model demand information according to the similarity between the demand text vector of the model demand information and the model text vector of the candidate model. Candidate models with higher similarity may be returned to the user as recommendations. Therefore, the method provided by the embodiment of the invention has a better model recommendation effect.
In addition, according to the method provided by the embodiment of the application, even if the user does not generate the behavior in the system, the user can be recommended with the desired candidate model according to the similarity between the demand text vector and the model text vector, and model recommendation is not needed according to the historical behavior of the user. Therefore, the problem of cold start can be solved well.
Further effects of the above-mentioned non-conventional alternatives will be described below in connection with the embodiments.
Drawings
The drawings are included to provide a better understanding of the invention and are not to be construed as unduly limiting the invention. Wherein:
FIG. 1 is a diagram illustrating an application scenario of a model recommendation method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating a flow of a model recommendation method according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating a flow of a method for generating a requirement text vector according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a vector space model building process according to an embodiment of the present invention;
FIG. 5 is a schematic diagram illustrating a flow of a method for generating model text vectors according to an embodiment of the present invention;
FIG. 6 is a schematic structural diagram of a model recommendation apparatus according to an embodiment of the present invention;
fig. 7 is a schematic block diagram of a computer system suitable for use in implementing a terminal device or server of an embodiment of the invention.
Detailed Description
Exemplary embodiments of the present invention are described below with reference to the accompanying drawings, in which various details of embodiments of the invention are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
In the process of recommending the model, screening can be directly carried out according to the keywords. The key word screening mode is premised on that the user has very definite models required by the user. The keyword screening is an accurate matching, but often, a user only has clear effects and actions to be achieved on the model, and it is unclear what algorithm is specifically used, how the model is built, and what operation is performed on the model for similar problems in the model library. The user often has clear demand information on the model, and wants to quickly find a similar and effective model from a large number of existing model research results.
Based on the above, the embodiment of the invention provides a model recommendation method with a better recommendation effect under the condition that a user is unclear about a model algorithm and a modeling method. Fig. 1 is a schematic diagram of an application scenario of a model recommendation method according to an embodiment of the present invention. As shown in fig. 1, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
The user sends a model recommendation request through the terminal devices 101, 102, and 103, where the model recommendation request may include a model name, a model classification, model text information, and the like. The terminal devices 101, 102, 103 may be cell phones, notebooks, servers, tablets, laptop portable computers, etc.
The server 105 determines the model recommendation requests sent by the terminal devices 101, 102 and 103; determining a demand text vector corresponding to the model demand information; and determining a recommendation result corresponding to the model demand information according to the similarity value between the demand text vector and the model text vector.
It should be noted that the model recommendation method provided by the embodiment of the present invention is generally executed by the server 105, and accordingly, the model recommendation apparatus is generally disposed in the server 105.
It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
Fig. 2 is a schematic diagram of a flow of a model recommendation method according to an embodiment of the present invention, as shown in fig. 2, the method includes:
step 201: determining model requirement information, wherein the model requirement information comprises at least one of the following: model name, model classification, and model text information.
The model recommendation request can be received first, and the model requirement information can be determined according to the model recommendation request. Specifically, the model requirement information can be determined by screening and combining the information of the model requirement form filled by the user.
The model text information not only contains model requirement link information, but also includes model development link information, model evaluation information, specific model package and other model related information, such as model functions, model effects, model algorithms, evaluation information and the like.
Step 202: and determining a demand text vector corresponding to the model demand information.
The model requirement information can be firstly abstracted into a piece of text information, namely the model requirement text. And integrating the model requirement information together, wherein each model requirement is represented by a model document.
And determining a demand text vector corresponding to the model demand information according to the model demand text. In the process of refining the model requirement text, the information which is relatively high in importance and aims at the model name and the model classification short text can be corrected by adopting a resampling method so as to strengthen the importance of the short text.
The method can be used for determining a demand text vector corresponding to the model demand information according to the model demand text:
preprocessing the model requirement text to generate a preprocessed text;
determining the weight of each term in the preprocessed text;
and generating a demand text vector corresponding to the model demand information according to the weight of the lexical item.
The pre-treatment may include at least one of: word segmentation processing, stop word filtering processing and synonym conversion processing. The preprocessed text is the text preprocessed for the model demand text, and the demand text vector corresponding to the model demand information can be conveniently generated according to the preprocessed text.
Step 203: for each candidate model: determining a model text vector corresponding to the candidate model; a similarity value between the demand text vector and the model text vector is determined.
And converting the model requirement information into a requirement text vector, and sequentially determining similarity values between the requirement text vector and each candidate model text vector in the model library. The similarity value may be calculated in a number of ways. For example, the euclidean distance, hamming distance, cosine similarity, etc. between the demand text vector and the model text vector are calculated.
Step 204: and determining a recommendation result corresponding to the model demand information according to the similarity value.
Selecting a preset number of candidate models from the candidate models according to the similarity value, wherein the similarity value between the text vector of the candidate models and the required text vector is the highest; and generating a recommendation result corresponding to the model demand information according to the preset number of candidate models.
In the embodiment of the invention, the recommendation result corresponding to the model demand information is determined according to the similarity between the demand text vector of the model demand information and the model text vector of the candidate model. The candidate models with higher similarity can be returned to the user as recommendation results, so that the model recommendation method has a better model recommendation effect. Further, the method of the embodiment of the application does not need to perform model recommendation according to the historical behaviors of the user. Therefore, the problem of cold start can be solved well.
In addition, the embodiment of the invention utilizes text similarity to carry out similarity model matching and recommendation. According to the requirement link information of the model, similar models are recommended from the model library, and compared with the traditional accurate screening, more suitable models can be found and given to the user.
Fig. 3 is a schematic diagram of a flow of a method for generating a requirement text vector according to an embodiment of the present invention, as shown in fig. 3, the method includes:
step 301: and generating a model requirement text corresponding to the model requirement information.
Step 302: and performing word segmentation on the model requirement text.
Word segmentation is a process of recombining continuous word sequences into word sequences according to a certain specification. And the Jieba word segmentation can be utilized to segment the model text information.
Step 303: and performing stop word filtering processing on the model requirement text after word segmentation processing.
The stop word means that some characters or words are automatically filtered before or after processing natural language data in the information retrieval process to save storage space and improve search efficiency. Filtering may be performed according to a list of commonly used stop words. Since the model descriptive text does not contain too many non-mainstream words such as emoticons, links, etc. Therefore, the stop word list adopts contents such as common English stop words, Chinese stop words, specific text segments with few meanings and the like.
Step 304: and carrying out synonym conversion processing on the filtered model requirement text to generate a preprocessed text.
Because the expression modes of some words in the common words described by the machine learning model are diversified but the words have the same meaning, in order to avoid the problem of matrix sparsity of a word vector model constructed subsequently, the embodiment of the invention carries out synonymy conversion operation on some common algorithm names, namely the algorithm names used by the model have Chinese, English names, short names and common names, and once the text encounters the words with different expressions of the same algorithm, the embodiment of the invention can map the words to the Chinese names with unified space and algorithm standards. The name of the algorithm is only one of common synonyms, and a specific user can supplement the synonym conversion table according to actual conditions.
Step 305: from the preprocessed text, a plurality of target terms is determined.
All terms in the preprocessed text may be targeted terms. And selecting a plurality of terms with higher importance degrees from all terms contained in the preprocessed text as target terms.
Step 306: and respectively determining the weight of each target term by utilizing a tf-dif algorithm.
DF (Document Frequency) refers to the number of documents containing a specified term. The formula of IDF (Inverse text Frequency index) is defined as follows:
where IDF (i) refers to the inverse document frequency of term i, N refers to the total number of documents in the model document set, and DF (i) refers to the document frequency of term i in the document set. It can be seen that the higher the document frequency of a term is, the lower the inverse document frequency of the term is, in other words, the more times a word appears in a document, the lower the discrimination of the term is, and the less obvious the feature is.
Wherein TF-dif is an index comprehensively considering Term TF (Frequency) and inverse document Frequency IDF, and the formula is as follows:
TFIDF(t,d)＝TF(t)·IDF(t,d)
TFIDF (t, d) refers to the weight of term t in document d. Tf (t) refers to the number of terms t in all model documents, and IDF (t, d) refers to the inverse document frequency of terms. TFIDF (t, d) is the weight of the target term.
If a term appears in all model documents, the weight of the term in the document set is lower; conversely, if a term appears multiple times in a small number of model documents, the term has a higher weight. The mechanism of comprehensively considering the term frequency and the inverse document frequency can well find the characteristic terms of the text.
Step 307: and determining a characteristic word set of the predicted text according to the weight of the terms, wherein the characteristic word set comprises at least one characteristic term.
There are various methods for determining the feature word set. For example, all target terms whose weights of terms exceed a preset weight may be determined as feature words, or a target term with the highest weight of a preset number of terms may be selected as a feature word.
Step 308: and acquiring a preset keyword set.
Step 309: and generating a demand text vector corresponding to the model demand information according to the feature word set and the keyword set.
A text vector is a structure that converts the actual text content into a structure that can be recognized inside the machine. And representing the required information of each model according to the characteristic words selected by the term weight to form a word document matrix. Fig. 4 is a schematic diagram of a construction process of a vector space model according to an embodiment of the present invention. Wherein, the keywords T1-TN in the keyword set are derived from the model textThe first n words extracted by the word frequency weight in the document set are T1-TN which is the keyword set. DiRepresented is the ith model requirement document, DiThe feature word set required by the ith model. WmnFor characterizing documents DmWhether or not to include the word TnIf D ismWhether or not to include the word TnContains, then WmnIs 1, whereas WmnIs 0. Wm1-WmnNamely, the requirement text vector corresponding to the mth requirement model information.
In an embodiment of the invention, a method for generating a requirement text vector from model requirement information is provided. According to the method, the characteristic words contained in the model demand information are extracted, so that the finally generated demand text vector can well represent the model demand information required by the user. In addition, synonym conversion processing is carried out on terms in the model requirement text, distribution of model vectors can be corrected, and meanings of the model vectors are enriched.
In order to speed up the calculation of the similarity value between the demand text vector and the model text vector and reduce unnecessary repeated calculation, the model text vector corresponding to each candidate model can be calculated in advance in the system. Fig. 5 is a schematic diagram of a flow of a model text vector generation method according to an embodiment of the present invention, and as shown in fig. 5, the method includes:
step 501: obtaining model description information of the candidate model, wherein the model description information comprises: model attribute information and model package information.
Optionally, the model attribute information comprises at least one of: model name, model classification, model output parameters, model input parameters, application objects and evaluation indexes. The model packet information includes a training code, a data processing code, a model prediction code, and the like of the model. Therefore, the text information of the model package is large.
Step 502: from the model package information, a plurality of terms are extracted.
Step 503: weights of the plurality of terms are determined, respectively.
The method for determining the weight of the term may refer to steps 301 to 306 in fig. 3, and is not described in detail.
Step 504: at least one keyword term of the model package is determined according to the weight of the term.
There are various methods for determining the feature word set. For example, all target terms whose weights of terms exceed a preset weight may be determined as feature words, or a target term with the highest weight of a preset number of terms may be selected as a feature word.
Step 505: and generating a model description text corresponding to the model description information according to the at least one keyword item and the model attribute information.
And fusing or simply splicing the key terms extracted from the model packet and the model attribute information to generate a model description text corresponding to the model description information. The fusion mode is that different weights are distributed to the key terms extracted from the model packet and the model attribute information so as to reflect different importance degrees between the key terms and the attribute information.
Step 506: and determining a model text vector corresponding to the candidate model according to the model description text.
The method for generating the model text vector can refer to steps 301 to 309 in fig. 3, and is not described in detail.
In an embodiment of the invention, a method for generating model text vectors from model description information is provided. According to the method, the feature words contained in the model description information are extracted, so that the finally generated model text vector can well represent the attribute information of each candidate model.
In addition, since there are many text messages in the model packet, if the text messages are directly spliced with other model attribute information, the distribution of the model information text vector may be unbalanced, and then before constructing the model text vector, keywords in the model packet need to be extracted according to the weights of terms to serve as texts in the model packet, and then the extracted texts participate in the fusion of other model attribute information, so as to construct the model text vector.
Fig. 6 is a schematic structural diagram of a model recommendation apparatus according to an embodiment of the present invention. As shown in fig. 6, the apparatus includes:
an information determining module 601, configured to determine model requirement information, where the model requirement information includes at least one of: model name, model classification and model text information;
a vector determination module 602, configured to determine a requirement text vector corresponding to the model requirement information;
a similarity value determination module 603 configured to, for each candidate model: determining a model text vector corresponding to the candidate model; determining a similarity value between the demand text vector and the model text vector;
and a result determining module 604, configured to determine, according to the similarity value, a recommendation result corresponding to the model requirement information.
Optionally, the information determining module 601 is specifically configured to:
receiving a model recommendation request;
and determining model demand information according to the model recommendation request.
Optionally, the vector determining module 602 is specifically configured to:
generating a model demand text corresponding to the model demand information;
and determining a demand text vector corresponding to the model demand information according to the model demand text.
Optionally, the vector determining module 602 is specifically configured to:
preprocessing the model requirement text to generate a preprocessed text;
determining the weight of each term in the preprocessed text;
and generating a demand text vector corresponding to the model demand information according to the weight of the lexical item.
Optionally, the vector determining module 602 is specifically configured to:
performing word segmentation on the model requirement text;
carrying out stop word filtering processing on the model requirement text after word segmentation processing;
and carrying out synonym conversion processing on the filtered model requirement text to generate a preprocessed text.
Optionally, the vector determining module 602 is specifically configured to:
determining a plurality of target terms from the preprocessed text;
and respectively determining the weight of each target term by utilizing a tf-dif algorithm.
Optionally, the vector determining module 602 is specifically configured to:
determining a characteristic word set of the predicted text according to the weight of the lexical item, wherein the characteristic word set comprises at least one characteristic lexical item;
acquiring a preset keyword set;
and generating a demand text vector corresponding to the model demand information according to the feature word set and the keyword set.
Optionally, the result determining module 604 is specifically configured to:
selecting a preset number of candidate models from the candidate models according to the similarity value;
and generating a recommendation result corresponding to the model demand information according to the preset number of candidate models.
Optionally, the apparatus further comprises:
and a vector generation module 605, configured to generate a model text vector corresponding to the candidate model.
Optionally, the vector generation module 605 is specifically configured to:
obtaining model description information of the candidate model;
generating a model description text corresponding to the model description information;
and determining a model text vector corresponding to the candidate model according to the model description text.
Optionally, the model description information includes: model attribute information and model package information;
the vector generation module 605 is specifically configured to:
extracting a plurality of terms from the model packet information;
determining weights of a plurality of terms respectively;
determining at least one keyword item of the model packet according to the weight of the keyword item;
and generating a model description text corresponding to the model description information according to the at least one keyword item and the model attribute information.
Optionally, the model attribute information comprises at least one of: model name, model classification, model output parameters, model input parameters, application objects and evaluation indexes.
An embodiment of the present invention provides an electronic device, including:
one or more processors;
a storage device for storing one or more programs,
when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method of any of the embodiments described above.
Referring now to FIG. 7, shown is a block diagram of a computer system 700 suitable for use with a terminal device implementing an embodiment of the present invention. The terminal device shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present invention.
As shown in fig. 7, the computer system 700 includes a Central Processing Unit (CPU)701, which can perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)702 or a program loaded from a storage section 708 into a Random Access Memory (RAM) 703. In the RAM 703, various programs and data necessary for the operation of the system 700 are also stored. The CPU 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.
The following components are connected to the I/O interface 705: an input portion 706 including a keyboard, a mouse, and the like; an output section 707 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 708 including a hard disk and the like; and a communication section 709 including a network interface card such as a LAN card, a modem, or the like. The communication section 709 performs communication processing via a network such as the internet. A drive 710 is also connected to the I/O interface 705 as needed. A removable medium 711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 710 as necessary, so that a computer program read out therefrom is mounted into the storage section 708 as necessary.
In particular, according to the embodiments of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 709, and/or installed from the removable medium 711. The computer program performs the above-described functions defined in the system of the present invention when executed by the Central Processing Unit (CPU) 701.
It should be noted that the computer readable medium shown in the present invention can be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present invention, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present invention, however, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The modules described in the embodiments of the present invention may be implemented by software or hardware. The described modules may also be provided in a processor, which may be described as: the device comprises an information determining module, a vector determining module, a similarity value determining module and a result determining module. Where the names of these modules do not in some cases constitute a limitation on the module itself, for example, the information determination module may also be described as a "module that determines model requirement information".
As another aspect, the present invention also provides a computer-readable medium that may be contained in the apparatus described in the above embodiments; or may be separate and not incorporated into the device. The computer readable medium carries one or more programs which, when executed by a device, cause the device to comprise:
determining model requirement information, the model requirement information comprising at least one of: model name, model classification and model text information;
determining a demand text vector corresponding to the model demand information;
for each candidate model: determining a model text vector corresponding to the candidate model; determining a similarity value between the demand text vector and the model text vector;
and determining a recommendation result corresponding to the model demand information according to the similarity value.
According to the technical scheme of the embodiment of the invention, the recommendation result corresponding to the model demand information is determined by utilizing the similarity between the demand text vector of the model demand information and the model text vector of the candidate model. The candidate models with higher similarity can be returned to the user for reference, so that the model recommendation method has a better model recommendation effect. In addition, model recommendation is not needed according to the historical behaviors of the user, and even if the user does not generate behaviors in the system, the user can be recommended with the desired candidate models according to the similarity.
The above-described embodiments should not be construed as limiting the scope of the invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions can occur, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (15)
1. A method for model recommendation, comprising:
determining model requirement information, the model requirement information comprising at least one of: model name, model classification and model text information;
determining a demand text vector corresponding to the model demand information;
for each candidate model: determining a model text vector corresponding to the candidate model; determining a similarity value between the demand text vector and the model text vector;
and determining a recommendation result corresponding to the model demand information according to the similarity value.
2. The method of claim 1, wherein the determining model requirement information comprises:
receiving a model recommendation request;
and determining the model demand information according to the model recommendation request.
3. The method of claim 1, wherein the determining the requirement text vector corresponding to the model requirement information comprises:
generating a model demand text corresponding to the model demand information;
and determining a demand text vector corresponding to the model demand information according to the model demand text.
4. The method according to claim 3, wherein the determining, according to the model requirement text, a requirement text vector corresponding to the model requirement information includes:
preprocessing the model requirement text to generate a preprocessed text;
determining the weight of each term in the preprocessed text;
and generating a demand text vector corresponding to the model demand information according to the weight of the lexical item.
5. The method of claim 4, wherein preprocessing the model requirement text to generate preprocessed text comprises:
performing word segmentation processing on the model requirement text;
carrying out stop word filtering processing on the model requirement text after word segmentation processing;
and carrying out synonym conversion processing on the filtered model requirement text to generate the preprocessed text.
6. The method of claim 4, wherein determining the weight of each term in the preprocessed text comprises:
determining a plurality of target terms from the preprocessed text;
and respectively determining the weight of each target term by utilizing a tf-dif algorithm.
7. The method of claim 4, wherein generating the requirement text vector corresponding to the model requirement information according to the weight of the term comprises:
determining a feature word set of the predicted text according to the weight of the lexical item, wherein the feature word set comprises at least one feature lexical item;
acquiring a preset keyword set;
and generating a demand text vector corresponding to the model demand information according to the feature word set and the keyword set.
8. The method according to claim 1, wherein the determining the recommendation result corresponding to the model requirement information according to the similarity value includes:
selecting a preset number of candidate models from the candidate models according to the similarity value;
and generating a recommendation result corresponding to the model demand information according to the preset number of candidate models.
9. The method of claim 1, wherein before determining the model text vector corresponding to the candidate model, further comprising:
and generating a model text vector corresponding to the candidate model.
10. The method of claim 9, wherein generating a model text vector corresponding to the candidate model comprises:
obtaining model description information of the candidate model;
generating a model description text corresponding to the model description information;
and determining a model text vector corresponding to the candidate model according to the model description text.
11. The method of claim 10, wherein the model description information comprises: model attribute information and model package information;
the generating of the model description text corresponding to the model description information includes:
extracting a plurality of terms from the model packet information;
respectively determining the weight of each term;
determining at least one keyword item of the model packet according to the weight of the keyword item;
and generating a model description text corresponding to the model description information according to the at least one keyword and the model attribute information.
12. The method of claim 11, wherein the model attribute information comprises at least one of: model name, model classification, model output parameters, model input parameters, application objects and evaluation indexes.
13. A model recommendation device, comprising:
an information determination module configured to determine model requirement information, the model requirement information including at least one of: model name, model classification and model text information;
the vector determination module is used for determining a demand text vector corresponding to the model demand information;
a similarity value determination module for, for each candidate model: determining a model text vector corresponding to the candidate model; determining a similarity value between the demand text vector and the model text vector;
and the result determining module is used for determining a recommendation result corresponding to the model demand information according to the similarity value.
14. An electronic device, comprising:
one or more processors;
a storage device for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-12.
15. A computer-readable medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1-12.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110338638.6A CN112926295A (en) | 2021-03-30 | 2021-03-30 | Model recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110338638.6A CN112926295A (en) | 2021-03-30 | 2021-03-30 | Model recommendation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112926295A true CN112926295A (en) | 2021-06-08 |
Family
ID=76176532
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110338638.6A Pending CN112926295A (en) | 2021-03-30 | 2021-03-30 | Model recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112926295A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117389544A (en) * | 2023-12-13 | 2024-01-12 | 北京宇信科技集团股份有限公司 | Artificial intelligence data modeling method, device, medium and equipment |
-
2021
- 2021-03-30 CN CN202110338638.6A patent/CN112926295A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117389544A (en) * | 2023-12-13 | 2024-01-12 | 北京宇信科技集团股份有限公司 | Artificial intelligence data modeling method, device, medium and equipment |
| CN117389544B (en) * | 2023-12-13 | 2024-03-01 | 北京宇信科技集团股份有限公司 | Artificial intelligence data modeling method, device, medium and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107491534B (en) | Information processing method and device | |
| CN107491547B (en) | Search method and device based on artificial intelligence | |
| CN107220386B (en) | Information pushing method and device | |
| Ding et al. | Entity discovery and assignment for opinion mining applications | |
| CN106960030B (en) | Information pushing method and device based on artificial intelligence | |
| CN110457708B (en) | Vocabulary mining method and device based on artificial intelligence, server and storage medium | |
| CN114861889B (en) | Deep learning model training method, target object detection method and device | |
| CN110147425A (en) | A kind of keyword extracting method, device, computer equipment and storage medium | |
| CN112926308B (en) | Method, device, equipment, storage medium and program product for matching text | |
| CN107526718A (en) | Method and apparatus for generating text | |
| CN113326420A (en) | Question retrieval method, device, electronic equipment and medium | |
| CN116028618B (en) | Text processing method, text searching method, text processing device, text searching device, electronic equipment and storage medium | |
| CN113268560A (en) | Method and device for text matching | |
| CN112686053A (en) | Data enhancement method and device, computer equipment and storage medium | |
| CN111813993A (en) | Video content expanding method and device, terminal equipment and storage medium | |
| CN111382563A (en) | Text relevance determining method and device | |
| WO2010132062A1 (en) | System and methods for sentiment analysis | |
| CN112926295A (en) | Model recommendation method and device | |
| CN112329429A (en) | Text similarity learning method, device, equipment and storage medium | |
| CN111126073A (en) | Semantic retrieval method and device | |
| CN112925872A (en) | Data searching method and device | |
| CN116048463A (en) | Intelligent recommendation method and device for content of demand item based on label management | |
| CN110895655A (en) | Method and device for extracting text core phrase | |
| CN111368036B (en) | Method and device for searching information | |
| CN112528644B (en) | Entity mounting method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |