Connect public, paid and private patent data with Google Patents Public Datasets

Entity assessment and ranking

Download PDF

Info

Publication number
US20140095466A1
US20140095466A1 US14097976 US201314097976A US20140095466A1 US 20140095466 A1 US20140095466 A1 US 20140095466A1 US 14097976 US14097976 US 14097976 US 201314097976 A US201314097976 A US 201314097976A US 20140095466 A1 US20140095466 A1 US 20140095466A1
Authority
US
Grant status
Application
Patent type
Prior art keywords
documents
set
entity
candidate
metadata
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.)
Abandoned
Application number
US14097976
Inventor
Chad Michael Cumby
Katherine Probst
Rayid Ghani
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Accenture Global Services Ltd
Original Assignee
Accenture Global Services Ltd
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

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRICAL DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/30Information retrieval; Database structures therefor ; File system structures therefor
    • G06F17/30286Information retrieval; Database structures therefor ; File system structures therefor in structured data stores
    • G06F17/30386Retrieval requests
    • G06F17/30424Query processing
    • G06F17/30522Query processing with adaptation to user needs
    • G06F17/3053Query processing with adaptation to user needs using ranking
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRICAL DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/30Information retrieval; Database structures therefor ; File system structures therefor
    • G06F17/3061Information retrieval; Database structures therefor ; File system structures therefor of unstructured textual data
    • G06F17/30634Querying
    • G06F17/30657Query processing
    • G06F17/3066Query translation
    • G06F17/30672Query expansion
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRICAL DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/30Information retrieval; Database structures therefor ; File system structures therefor
    • G06F17/3061Information retrieval; Database structures therefor ; File system structures therefor of unstructured textual data
    • G06F17/30634Querying
    • G06F17/30657Query processing
    • G06F17/30675Query execution
    • G06F17/30687Query execution using probabilistic model

Abstract

General entity retrieval and ranking is described. A first set of documents is retrieved from one or more document repositories based on a query formed according to the topic. The first set of documents is characterized based on its first set of metadata values. One or more candidate entities are identified based on the first set of documents and the original query is thereafter augmented according to a candidate entity. The second set of documents resulting from the augmented query is then characterized in a similar manner. For each candidate entity, the first and second document set characterizations are compared to determine their degree of similarity. Increasingly similar document set characterizations indicates that the candidate entity is increasingly relevant to the original query. Repeating this process for each of the one or more candidate entities can give rise to rankings according to the respective degrees of similarity.

Description

    FIELD OF THE INVENTION
  • [0001]
    The instant disclosure relates generally to ranking systems and, in particular, to techniques for assessing and ranking entities.
  • BACKGROUND OF THE INVENTION
  • [0002]
    With the advent of the World Wide Web and Internet, the volume of publicly available information has grown at an unprecedented rate. In order to make sense of this ever-expanding collection, significant attention has been paid to the development of improved document searching techniques, such as search engines and the like. While such techniques have greatly improved the speed, cost and accuracy of locating relevant documents in an essentially unstructured knowledge base, the realm of entity retrieval and ranking, until recently, has been the subject of limited research. As used herein, an entity is defined by its ability to be described by one or more nouns, e.g., a person, place or thing. By way of non-limiting example, in the context of commercial enterprises, entities may comprise employees, clients, projects, partners, alliances, facility locations, competitors, etc. Of course, similar entities will be readily apparent in numerous endeavors beyond the commercial context. Regardless, the ability to quickly identify entities relevant to a given topic of interest will find application in a wide variety of applications.
  • [0003]
    For example, referring again to the commercial context, the preparation of business proposals may be made more efficient if one is able to quickly identify subject matter experts within the organization submitting the proposal. In a similar vein, the ability to accurately identify the most qualified potential team members with specific skill sets would improve project staffing. Further still, identifying the best vendors for certain equipment or service needs would be greatly simplified through provision of a system that enables quick and accurate identification of relevant entities. Stated more generally, various knowledge management tasks can be greatly simplified or assisted by delivering relevant information about entities to those responsible for such knowledge management tasks.
  • [0004]
    Currently, it is very difficult to retrieve entity-related information. In a business context, any commercial enterprise search engine, in a manner akin to web search engines, will yield a list of documents relevant to a particular topic query. However, such engines are of little help in retrieving a reliable ranked list of entities relevant to the topic, and it is left to the requestor to sift through the returned documents to identify any particularly relevant entities.
  • [0005]
    More recently, entity, and especially expert, ranking has received a growing amount of attention. For example, the Initiative for the Evaluation of XML Retrieval (INEX) has introduced an entity ranking track. Such systems currently rely on the retrieved entities being marked up with Extensible Markup Language (XML). However, not all content within a given knowledge base may have entities tagged with appropriate mark-up. The Text Retrieval Conference (TREC) recently introduced an enterprise track, including an expert finding task. In one approach, a list of experts is provided and, for a given expert, a pseudo-document is created from all documents located that include a mention of that expert. In another approach, potentially relevant documents for a topic are retrieved and experts are subsequently extracted from (i.e., identified in) the set of documents. Ranking of the extracted experts according to their relevance to the topic is inferred by the number of mentions for each expert; more mentions results in higher rankings. However, to the extent that the number of mentions of an expert in a set of documents is subject to numerous other factors beyond relevance to a given topic, such systems are susceptible to providing inaccurate results. Further still, some expert identification techniques exploit structural information of documents, such as references from other, topically relevant documents or, in the example of emails, explicit links to other emails. With regard to these expert identification techniques, expert retrieval, while important, is appropriately viewed as a subset of entity retrieval and ranking and is thus limited in scope. That is, a more general entity retrieval and ranking approach represents a more scalable solution allowing for application to a wider variety of situations, and would therefore represent an advancement in the art.
  • SUMMARY OF THE INVENTION
  • [0006]
    The instant disclosure describes techniques for general entity retrieval and ranking based on specific topic queries directed to document repositories. In particular, the instant disclosure describes techniques that leverage the availability of metadata about the documents being searched, which metadata is often more available in enterprise document repositories. The disclosed techniques may be implemented using suitable processing devices, such as general purpose or application specific computers, or other equivalent implementation techniques known in the art.
  • [0007]
    In one embodiment, a user may directly, or via an intervening component, provide a topic that is subsequently formed into a query. Based on the query, a first set of documents is retrieved from one or more document repositories, for example via a suitable search engine. The first set of documents have first metadata values for a corresponding plurality of metadata attributes. The first set of documents is then characterized based on the first set of metadata values. One or more candidate entities are then identified based on the first set of documents. For example, candidate entities may be identified through text-extraction applied to the first set of documents, the number of mentions in the first set of documents or directly from the first metadata values. In one embodiment, the one or more candidate entities are selected according to an entity type, potentially provided by the querying user.
  • [0008]
    Thereafter, the original query is augmented according to the one or more candidate entities. That is, document repository(ies) are searched again based on the original query and one of the candidate entities. The resulting second set of documents is then characterized on the basis of the same metadata attributes and the second metadata values associated with the second set of documents. In one embodiment, a document set is characterized by creating a vector in which each of the metadata values constitutes a separate dimension, optionally with weighting values for specific metadata values applied. Regardless, the first and second document set characterizations are then compared (e.g., through a vector comparison) to determine their degree of similarity. Increasingly similar document set characterizations lead to the inference that the candidate entity giving rise to the second document set is increasingly relevant to the original query. The intuition behind this metric is that the metadata values provide a detailed model of the documents in a retrieved set. Further, the metadata describes not only the content of each document (e.g., by automatic retrieval of named entities or keywords), but also other information associated with it. In other words, it describes the “essence” of the document along with all relevant data about it, e.g., its type, associated part of an organization, etc. Repeating this process for each of the one or more candidate entities can give rise to rankings according to the respective degrees of similarity, which rankings can be subsequently provided in ordered list form. In this manner, the disclosed techniques represent an advancement in the art.
  • BRIEF DESCRIPTION OF THE DRAWINGS
  • [0009]
    The features described in this disclosure are set forth with particularity in the appended claims. These features and attendant advantages will become apparent from consideration of the following detailed description, taken in conjunction with the accompanying drawings. One or more embodiments are now described, by way of example only, with reference to the accompanying drawings wherein like reference numerals represent like elements and in which:
  • [0010]
    FIG. 1 is a block diagram of a processing device the may be used to implement the various techniques described herein;
  • [0011]
    FIG. 2 is a flowchart of processing in accordance with the instant disclosure;
  • [0012]
    FIG. 3 is an illustration of metadata attributes and metadata values for a plurality of candidate entities in accordance with the instant disclosure; and
  • [0013]
    FIG. 4 is a block diagram of a system comprising an entity ranking component that may be used to implement the various techniques described herein.
  • DETAILED DESCRIPTION OF THE PRESENT EMBODIMENTS
  • [0014]
    FIG. 1 illustrates an exemplary processing device 100 that may be used to implement the teachings of the instant disclosure. The device 100 may be used to implement, for example, one or more components of the system 400, as described in greater detail below. Regardless, the device 100 comprises a processor 102 coupled to a storage component 104. The storage component 104, in turn, comprises stored executable instructions 116 and data 118. In an embodiment, the processor 102 may comprise one or more processing devices such as a microprocessor, microcontroller, digital signal processor, or combinations thereof capable of executing the stored instructions 116 and operating upon the stored data 118. Likewise, the storage component 104 may comprise one or more devices such as volatile or nonvolatile memory including but not limited to random access memory (RAM) or read only memory (ROM). Processor and storage arrangements of the types illustrated in FIG. 1 are well known to those having ordinary skill in the art. In one embodiment, the processing techniques described herein are implemented as a combination of executable instructions and data within the storage component 104.
  • [0015]
    As shown, the device 100 may comprise one or more user input devices 106, a display 108, a peripheral interface 110, other output devices 112 and a network interface 114 in communication with the processor 102. The user input device 106 may comprise any mechanism for providing user input (such as inputs specifying query topics, entity types, etc. as described below) to the processor 102. For example, the user input device 106 may comprise a keyboard, a mouse, a touch screen, microphone and suitable voice recognition application or any other means whereby a user of the device 100 may provide input data to the processor 102. The display 108, may comprise any conventional display mechanism such as a cathode ray tube (CRT), flat panel display, or any other display mechanism known to those having ordinary skill in the art. The peripheral interface 110 may include the hardware, firmware and/or software necessary for communication with various peripheral devices, such as media drives (e.g., magnetic disk or optical disk drives), other processing devices or any other input source used in connection with the instant techniques. Likewise, the other output device(s) 112 may optionally comprise similar media drive mechanisms, other processing devices or other output destinations capable of providing information to a user of the device 100, such as speakers, LEDs, tactile outputs, etc. Finally, the network interface 114 may comprise hardware, firmware and/or software that allows the processor 102 to communicate with other devices via wired or wireless networks, whether local or wide area, private or public, as known in the art. For example, such networks may include the World Wide Web or Internet, or private enterprise networks, as known in the art.
  • [0016]
    While the device 100 has been described as a one form for implementing the techniques described herein, those having ordinary skill in the art will appreciate that other, functionally equivalent techniques may be employed. For example, as known in the art, some or all of the executable instruction-implemented functionality may be implemented using firmware and/or hardware devices such as application specific integrated circuits (ASICs), programmable logic arrays, state machines, etc. Further still, other implementations of the device 100 may include a greater or lesser number of components than those illustrated. Once again, those of ordinary skill in the art will appreciate the wide number of variations that may be used is this manner.
  • [0017]
    Referring now to FIG. 2, processing in accordance with the instant disclosure is further described. While the processing illustrated in FIG. 2 may be implemented as stored instructions executed on one or more processors (as in FIG. 1), those of ordinary skill in the art will appreciate that other implementation techniques are possible. Regardless, processing begins at block 202 where information concerning a given topic, is received. In one embodiment described below relative to FIG. 4, the topic information is received directly from a user via an appropriate user interface. However, the instant disclosure is not limited in this regard; the topic information may be received from an intervening device (e.g., a suitable computer not implementing a user interface) configured to provide the topic input.
  • [0018]
    Regardless of the manner in which the topic is received, processing continues at block 204 where the topic is used as the basis for retrieving a first set of documents. In one embodiment, this is done by fashioning the topic into a query, qt, suitable for submission to an appropriate search engine. A typical query may be formatted as one or more keywords that may optionally include logical operators (e.g., “and”, “or”, “not”, etc.) as known in the art. The keywords may be taken directly from the received topic information or indirectly determined, e.g., a topic expressed in natural language form and parsed for keywords, as further known in the art. Based on the query, the search engine searches documents in one or more repositories for documents meeting the query. For example, the “FAST” enterprise search engine provided by Fast Search & Transfer may be used for this purpose, although other search engines may be equally employed.
  • [0019]
    Once the first set of documents, Rt, has been retrieved in response to the query, processing continues at block 206 where one or more candidate entities, ei, are identified based on the first set of documents. In one embodiment, entities annotated by the document metadata may be used for this purpose. Alternatively, conventional entity extraction techniques may be employed. For example, the “FAST” search engine noted above possesses a named entity recognition capability that may be used to extract various entities. In this example, the search engine indexes automatically extracted entities such as companies, people, keywords, locations, acronyms, etc. as well as manually provided entities such as project, client, project contact, etc. from the documents in the first set of documents. Those having ordinary skill in the art will appreciate that other algorithms may be employed for this purpose. Regardless of the manner in which they are identified, occurrence frequencies for each entity, cnte i , may be generated across the first set of documents. Based on the occurrence frequencies, only a top portion of the entities, i.e., those having the highest occurrence frequencies, are used as candidate entities for subsequent processing. This may accomplished by choosing only the top n candidate entities as ranked by the occurrence frequencies or by thresholding the candidate entities, i.e., selecting those candidate entities where cnte i ≧CountThreshold.
  • [0020]
    In one embodiment, an entity type may be specified prior to the entity identification process. Examples of entity types include, but are not limited to, “companies”, “people”, “projects”, “locations”, etc. Generally, the types of entities specified may depend on the nature of the topic. Regardless, the entity type may be specified by the user in the same manner used to specify the topic or in an automated fashion.
  • [0021]
    Having identified one or more candidate entities, ei, processing continues at block 208, where the first set of documents is characterized. Characterization of document sets, as used herein, summarizes the nature or makeup of the documents within the document set in a manner that facilitates comparisons with other document sets. To this end, in one embodiment, characterization of documents sets is carried out through the use of metadata associated with the documents in a document set. Often, metadata attributes (and their corresponding values) carry important information about each document, and in aggregation they carry enough information to describe the document base, or the subset of the collection that is described by a query.
  • [0022]
    One of the main differences between enterprise information retrieval and web retrieval is that in the enterprise context, each document in a collection typically has metadata associated therewith. While documents themselves may have few direct links to each other, the process of collection, storage, and authorship often creates a rich set of information on many attributes of the items in question. In the case, for example, of a project-based consulting organization, a set of submitted documents might be annotated with values for attributes about the project for which they were created. Such attributes could include the creator of a document, his/her employee level, specialization, location, the date/time the document was submitted, the client of the project, any vendors or technologies used, business contacts for the project, dates for objective completions, etc.
  • [0023]
    Typically, such metadata is entered at the time that a document is added to a content database or document repository. However, even if metadata is not captured at the time of data submission, off-the-shelf technologies are available to extract simple entities such as people, locations, organizations, and acronyms used. Regardless of the manner in which it is obtained, the instant techniques may performed using a relatively small number of metadata attributes and attribute values up to hundreds of attributes having dozens of values. For example, Table 1 illustrates metadata attributes and values associated with documents taken from a content database of a large consulting firm. The particular examples illustrated in FIG. 1 were drawn from a document set retrieved in response to the query “BP drilling”. The numbers in parentheses indicate the number of documents that are associated with the particular metadata value in the retrieved set.
  • [0000]
    TABLE 1
    Attribute Example Values
    Language Albanian, Arabic, Albanian, Arabic, Bulgarian, Catalan,
    Chinese (simple), Chinese (traditional), Finnish, Japanese,
    Polish, Russian, Slovak, Spanish, Swedish, Thai, Turkish
    Acronyms bim = business integration methodology
    erp = enterprise resource planning
    bpo = business process outsourcing
    edi = electronic data interchange
    crm = customer relationship management
    Companies ExxonMobil (137)
    Texaco (114)
    Statoil (110)
    Unocal (98)
    PeopleSoft (94)
    Keywords Supply Chain (134)
    Finance (61)
    Performance Management (61)
    Procure to Pay (54)
    Locations America (252)
    Canada (209)
    Asia (206) Houston (206)
    Texas (184)
    Industry Energy (301)
    Keywords Upstream (168) Utilities (29)
    Gas & Power (25)
    Government (23)
    Offerings Resources (188)
    Outsourcing & BPO (155)
    Human Performance (155)
    HR Transformation (115)
    Vendor SAP (55)
    Keywords Oracle (31)
    Alcatel (11)
    Plumtree (9)
  • [0024]
    In one embodiment, the characterization of document sets is performed by creating a multidimensional metadata vector, vt, that models Rt. Preferably, all metadata attributes associated with the documents (both manual and automatically extracted) are used in constructing the metadata vector, although some lesser quantity of the available metadata attributes could be employed. The metadata attributes used to construct the metadata vector can include the various entity categories described above, e.g., people or locations, as well as other, non-entity based metadata associated with the documents. For each metadata attribute, the metadata vector comprises, for each identified metadata value, a count of the number of documents within the set associated with the given metadata value. An example of this is illustrated in FIG. 3. As shown in FIG. 3, a plurality of metadata attributes, Ai for i=1 to N includes all of the metadata attributes used to characterize the set of documents. For a given document set, the metadata vector comprises a plurality of counts, cij(k), as shown, where i designates a metadata attribute, j indexes the various metadata values for the i′th attribute and k represents a particular document set. Thus, for the first set of documents (i.e., k=1), each attribute value has an associated count, cij(l) where j (the number of different values for a given i′th attribute) may vary from 0 to Ji(l), the number of different values for the i′th attribute in the first set of documents. As shown, the number of different values for each attribute in a given document set may vary. In one embodiment, each of the value counts in the metadata vector is normalized, i.e.,
  • [0000]
    v ij = c ij R t .
  • [0025]
    Referring once again to FIG. 2, having characterized the first document set, processing continues at block 210 where processing of the candidate entities to determine their ranking according to their relevance to the topic, t, begins. To this end, it is determined at block 210 whether any candidate entities remain to be processed. If so, processing continues at block 212 where the next candidate entity is retrieved and, at block 214, used as the basis in an augmented query based on the original topic and the candidate entity, i.e., t+ei, using the same query formation techniques described above. The resulting second set of documents is then characterized, at block 216, based on second metadata values corresponding to the second set of documents using the same techniques described above relative to block 208. Note that, in the multidimensional vector representation describe above, and as shown in FIG. 3, the number of values for a given attribute can vary between different documents, i.e., it is possible that Ji(km)≠Ji(kn). To account for this, the overall dimension of the multidimensional vector may be equal to the total number of all attribute values for all of the attributes. For any document sets not having a particular value for a given attribute, a zero count may be used such that the same dimensionality is employed for each document set.
  • [0026]
    Thereafter, at block 218, the first and second document set characterizations are compared to determine their corresponding degree of similarity. In the case of the multidimensional vector representations, the comparison may be carried out as a vector comparison. In particular, in accordance with well known techniques, the cosine distance between the vectors for the first and second document set characterizations may be used:
  • [0000]
    d ( v t + e i , v t ) = 1 v t , 1 * v t + e i , 1 v t * v t + e i Eq . 1
  • [0027]
    Compared in this manner, smaller distances between the vectors correspond to greater similarity between the first and second document sets. In turn, greater similarity between first and second document sets leads to the inference that the candidate entity, ei, giving rise to the second document set has increased relevance to the topic, t. Note that, although not shown in Equation 1, certain dimensions within the vectors, i.e., specific metadata values, could be weighted to have a greater or lesser effect on the distance calculation thereby making the relevancy determination more finely controlled. At block 220, a ranking for the candidate entity is determined according to the comparison result. In the case of the cosine distance of Equation 1, the ranking may be selected inversely relative to the distance, thereby capturing the intuition that smaller distances correspond to higher degrees of relevance.
  • [0028]
    The process of retrieving second documents sets, characterizing and comparing to the first document set is repeated for each of the one or more candidate entities as indicated. When no more candidate entities remain to be processed in this manner, processing continues at block 222 where an ordered listing according to the rankings for each of the candidate listings is provided. In the cosine distance embodiment, given the inverse relationship between distance and ranking, the candidate entity having the smallest distance, i.e., argmine i (d(vt+e i ,vt)), will be ranked highest. The resulting listing may be provided thereafter to the user (or requesting device).
  • [0029]
    Referring now to FIG. 4, a system 400 that includes an entity ranking component 410 is further described. As shown, the system 400 comprises the entity ranking component in communication with a search engine component 420 and a user interface component 430. The search engine component 420 communicates with a document/metadata store 440. One or more processing devices, e.g., the processing device 100 described above, may be used to implement the entity ranking component 410 and/or the user interface component 430. The user interface component 430, which may comprise a suitable graphical user interface as known in the art, may be directly coupled to the entity ranking component 410 as in the case where both components are implemented in the same processing device (e.g., a unified software application running on a desktop or laptop computer), or may communicate via one or more intervening networks as in the case where the user interface 430 is implemented in one processing device (e.g., an interface application running on a desktop or laptop computer) and the entity ranking component 410 is implemented in another processing device (e.g., a Web-based web application server). In a similar vein, the search engine component 420 may be implemented through either the same processing device or a different processing device, as the entity ranking component 410. The document/metadata store 440 comprises the one or more repositories noted above storing the documents to be searched and their associated metadata, and may be implemented using any suitable storage device(s), such as one or more appropriately programmed database servers or the like.
  • [0030]
    The entity ranking component 410 comprises a variety of sub-components for implementing various portions of the techniques described above. In particular, a query formation sub-component 412, responsive to topic and/or entity type inputs 432 received from the user interface 430, is provided to form the queries and augmented queries 426 and provide them to the search engine component 420. Once again, it is noted that the topic and/or entity type inputs 432 may be provided by a device other than the user interface 430. Information regarding the retrieved document sets and their aggregation of metadata 428, i.e., their characterizations, is provided by the search engine component 420 and placed in a retrieved documents store 414. Note that the metadata aggregation can be performed by a component other than the search engine, e.g., by the entity ranking component. The returned document information is used by the query formation sub-component 412 to formulate augmented queries, as noted above. A comparator sub-component 418 performs the similarity comparisons of the document set characterizations and provides the results to the ranking sub-component 419 that provides the listing of ranked entities 434, as described above.
  • [0031]
    As shown, the search engine component 420 includes a document search index 422 and metadata aggregator or characterization sub-component 424. In one embodiment, the document search index 422 is a data structure that stores information about words and metadata associated with documents, thereby permitting quick document query retrieval and ranking response. The metadata aggregator 424 performs the characterization process on the retrieved document sets. Furthermore, automatic tagging/entity extraction (i.e., the automated identification and tagging of entities in documents), if provided, may be performed in a variety of manners as shown in FIG. 4. In one embodiment, the automatic tagging/entity extraction 450 may be performed (as implemented by one or more suitable processing devices, which may include the processing devices used to implement the other components illustrated in FIG. 4) directly on the documents stored in the document/metadata store 440. Alternatively, the automatic tagging/entity extraction 416 may be optionally implemented as part of the entity ranking component 410 operating directly on any documents stored in retrieved document store 414.
  • [0032]
    The system 400 may be beneficially employed in a number of contexts where entity ranking would be advantageous. For example, in addition to the proposal writing application previously noted, in the context of project staffing, entity ranking can retrieve the expertise of any employee of an organization, even if this expertise was not explicitly specified by the employee. Using the above described techniques, expertise (expressed as a topic) can be derived by retrieving relevant information, such as clients for which the employee has worked, keywords that describe projects that the employee was involved in, locations in which the employee has been staffed, etc. Further still, a requesting user can also get a ranked list of people that are “related” to the employee in terms of their expertise and experience. This can be useful for identifying current and past supervisors, and identifying a list of other potential matches (employees) for a project. Yet another example application is vendor evaluation. Often, users do not have a good list of vendors to start from when attempting to source goods or services, causing them to spend considerable effort finding relevant vendors. Using the above-described techniques, at least an initial list of relevant vendors may be identified. These applications are but a few examples of the various applications that could benefit from the highly flexible entity ranking capabilities described herein.
  • [0033]
    While particular preferred embodiments have been shown and described, those skilled in the art will appreciate that changes and modifications may be made without departing from the instant teachings. It is therefore contemplated that any and all modifications, variations or equivalents of the above-described teachings fall within the scope of the basic underlying principles disclosed above and claimed herein.

Claims (21)

1-18. (canceled)
19. An entity determination method for a query, the method comprising:
determining characterizations of a first set of documents returned as search results for the query based on metadata associated with the first set of documents;
determining, by a processing device, candidate entities for the first set of documents;
determining a similarity of each candidate entity to the query based on the characterizations of the first set of documents and information associated with each candidate entity.
20. The method of claim 19, comprising:
determining a ranked order of the candidate entities according to the similarities determined for the candidate entities; and
presenting the ranked order.
21. The method of claim 19, wherein the determining of the information associated with each candidate entity comprises:
for each candidate entity,
retrieving a second set of documents based on a query for the candidate entity, wherein the second set of documents have associated metadata, and
determining characterizations of the second set of documents based on the second set of documents and the metadata associated with the second set of documents.
22. The method of claim 21, wherein the determining of the similarity of each candidate entity to the query comprises:
comparing the characterizations of the second set of documents to the characterizations of the first set of documents to determine the similarity for each candidate entity.
23. The method of claim 22, wherein the characterizations of the first and second set of documents comprise attribute vectors describing each of the documents in the first and second sets.
24. The method of claim 23, wherein the comparing of the characterizations of the second set of documents to the characterizations of the first set of documents comprises:
determining distances between the attribute vectors of the documents in the first and second sets; and
comparing the distances to determine the candidate entities similar to the query.
25. The method of claim 19, comprising:
receiving information identifying a topic;
generating the query from the information; and
executing the query on documents in the document repository.
26. The method of claim 25, wherein the determining of the similarity of each candidate entity to the query comprises determining a similarity of each candidate entity to the topic.
27. The method of claim 25, wherein the determining of the candidate entities for the first set of documents comprises determining top K candidate entities having a closest similarity to the topic, wherein K is an integer greater than 1.
28. The method of claim 27, wherein the determining of the top K candidate entities comprises:
determining occurrence frequencies for each identified candidate entity in the first set of documents; and
selecting the top K candidate entities based on the occurrence frequencies.
29. A search engine system comprising:
a data repository to store documents;
a document search engine executed by at least one processor to execute a query associated with a topic on the data repository;
a document characterizer to determine characterizations of a first set of documents returned as search results for the query based on metadata associated with the first set of documents; and
an entity determination subsystem to determine candidate entities for the first set of documents and determine a similarity of each candidate entity to the topic of the query based on the characterizations of the first set of documents and information associated with each candidate entity.
30. The search engine system of claim 29, wherein the entity determination subsystem is to determine a ranked order of the candidate entities according to the similarities determined for the candidate entities.
31. The search engine system of claim 29, comprising:
a user interface to present the ranked order to a user.
32. The search engine system of claim 29, wherein the entity determination subsystem is to determine the information associated with each candidate entity by:
for each candidate entity,
retrieving a second set of documents based on a query for the candidate entity, wherein the second set of documents have metadata, and
determining characterizations of the second set of documents based on the second set of documents and the metadata for the second set of documents.
33. The search engine system of claim 29, wherein to determine the similarity of each candidate entity to the topic, the entity determination subsystem is to compare the characterizations of the second set of documents to the characterizations of the first set of documents.
34. The search engine system of claim 33, wherein the characterizations of the first and second set of documents comprise attribute vectors describing each of the documents in the first and second sets.
35. The search engine system of claim 34, wherein to compare the characterizations of the second set of documents to the characterizations of the first set of documents, the entity determination subsystem is to determine distances between the attribute vectors of the documents in the first and second sets.
36. A system to determine entities associated with a topic based on query search results, the system comprising:
an interface to receive information associated with a topic from a user;
an entity ranking subsystem executed by at least one processor to determine characterizations of a first set of documents returned as search results for the query based on metadata associated with the first set of documents, determine candidate entities for the first set of documents, determine a similarity of each candidate entity to the topic of the query based on the characterizations of the first set of documents and information associated with each candidate entity, and determine rankings of the candidate entities based on their similarity to the topic,
wherein the entity ranking subsystem is to present the rankings via the interface.
37. The system of claim 36, wherein the entity ranking subsystem is to, for each candidate entity,
retrieve a second set of documents based on a query for the candidate entity, wherein the second set of documents have metadata, and
determine characterizations of the second set of documents based on the second set of documents and the metadata for the second set of documents.
38. The system of claim 36, wherein the entity ranking subsystem is to compare the characterizations of the second set of documents to the characterizations of the first set of documents to determine the similarity for each candidate entity.
US14097976 2008-12-29 2013-12-05 Entity assessment and ranking Abandoned US20140095466A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US12344738 US8639682B2 (en) 2008-12-29 2008-12-29 Entity assessment and ranking
US14097976 US20140095466A1 (en) 2008-12-29 2013-12-05 Entity assessment and ranking

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US14097976 US20140095466A1 (en) 2008-12-29 2013-12-05 Entity assessment and ranking

Related Parent Applications (1)

Application Number Title Priority Date Filing Date
US12344738 Continuation US8639682B2 (en) 2008-12-29 2008-12-29 Entity assessment and ranking

Publications (1)

Publication Number Publication Date
US20140095466A1 true true US20140095466A1 (en) 2014-04-03

Family

ID=42286178

Family Applications (2)

Application Number Title Priority Date Filing Date
US12344738 Active 2031-10-23 US8639682B2 (en) 2008-12-29 2008-12-29 Entity assessment and ranking
US14097976 Abandoned US20140095466A1 (en) 2008-12-29 2013-12-05 Entity assessment and ranking

Family Applications Before (1)

Application Number Title Priority Date Filing Date
US12344738 Active 2031-10-23 US8639682B2 (en) 2008-12-29 2008-12-29 Entity assessment and ranking

Country Status (2)

Country Link
US (2) US8639682B2 (en)
CA (1) CA2688642A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140136504A1 (en) * 2012-11-12 2014-05-15 Microsoft Corporation Multidimensional search architecture

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1889386A1 (en) * 2005-06-07 2008-02-20 Nokia Corporation Receiving devices
US8707262B2 (en) * 2011-12-22 2014-04-22 Sap Ag Code scoring
US8997008B2 (en) * 2012-07-17 2015-03-31 Pelicans Networks Ltd. System and method for searching through a graphic user interface
US9047278B1 (en) 2012-11-09 2015-06-02 Google Inc. Identifying and ranking attributes of entities
US9262510B2 (en) 2013-05-10 2016-02-16 International Business Machines Corporation Document tagging and retrieval using per-subject dictionaries including subject-determining-power scores for entries
US9251136B2 (en) * 2013-10-16 2016-02-02 International Business Machines Corporation Document tagging and retrieval using entity specifiers
US9235638B2 (en) 2013-11-12 2016-01-12 International Business Machines Corporation Document retrieval using internal dictionary-hierarchies to adjust per-subject match results

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6018738A (en) * 1998-01-22 2000-01-25 Microsft Corporation Methods and apparatus for matching entities and for predicting an attribute of an entity based on an attribute frequency value
US6144964A (en) * 1998-01-22 2000-11-07 Microsoft Corporation Methods and apparatus for tuning a match between entities having attributes
US6345264B1 (en) * 1998-01-22 2002-02-05 Microsoft Corporation Methods and apparatus, using expansion attributes having default, values, for matching entities and predicting an attribute of an entity
US20050149538A1 (en) * 2003-11-20 2005-07-07 Sadanand Singh Systems and methods for creating and publishing relational data bases
US20050154713A1 (en) * 2004-01-14 2005-07-14 Nec Laboratories America, Inc. Systems and methods for determining document relationship and automatic query expansion
US20050234881A1 (en) * 2004-04-16 2005-10-20 Anna Burago Search wizard
US20050289100A1 (en) * 2004-06-25 2005-12-29 International Business Machines Corporation Techniques for representing relationships between queries
US20060242130A1 (en) * 2005-04-23 2006-10-26 Clenova, Llc Information retrieval using conjunctive search and link discovery
US20070038610A1 (en) * 2001-06-22 2007-02-15 Nosa Omoigui System and method for knowledge retrieval, management, delivery and presentation
US20080140616A1 (en) * 2005-09-21 2008-06-12 Nicolas Encina Document processing
US20080189269A1 (en) * 2006-11-07 2008-08-07 Fast Search & Transfer Asa Relevance-weighted navigation in information access, search and retrieval
US20080306914A1 (en) * 2007-06-05 2008-12-11 Search Capital Ltd Method and system for performing a search
US7792858B2 (en) * 2005-12-21 2010-09-07 Ebay Inc. Computer-implemented method and system for combining keywords into logical clusters that share similar behavior with respect to a considered dimension
US7933904B2 (en) * 2007-04-10 2011-04-26 Nelson Cliff File search engine and computerized method of tagging files with vectors

Family Cites Families (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2184518A1 (en) 1996-08-30 1998-03-01 Jim Reed Real time structured summary search engine
US5943670A (en) * 1997-11-21 1999-08-24 International Business Machines Corporation System and method for categorizing objects in combined categories
US7630986B1 (en) * 1999-10-27 2009-12-08 Pinpoint, Incorporated Secure data interchange
US6397215B1 (en) 1999-10-29 2002-05-28 International Business Machines Corporation Method and system for automatic comparison of text classifications
US6496838B1 (en) 1999-12-31 2002-12-17 Qwest Communications International Inc. Database reconciliation method and system
FR2819326B1 (en) 2001-01-08 2003-03-21 Aloa Technologies Method for identifying an individual by reading a card and means for implementing such process
US7644144B1 (en) * 2001-12-21 2010-01-05 Microsoft Corporation Methods, tools, and interfaces for the dynamic assignment of people to groups to enable enhanced communication and collaboration
US7343302B2 (en) * 2002-02-11 2008-03-11 Puresense Environmental Inc. System and method for emergency response
US20040006559A1 (en) * 2002-05-29 2004-01-08 Gange David M. System, apparatus, and method for user tunable and selectable searching of a database using a weigthted quantized feature vector
US20060074980A1 (en) * 2004-09-29 2006-04-06 Sarkar Pte. Ltd. System for semantically disambiguating text information
US7716198B2 (en) * 2004-12-21 2010-05-11 Microsoft Corporation Ranking search results using feature extraction
CA2609916A1 (en) 2005-05-31 2006-12-07 Siemens Medical Solutions Usa, Inc. System and method for data sensitive filtering of patient demographic record queries
US20070118399A1 (en) * 2005-11-22 2007-05-24 Avinash Gopal B System and method for integrated learning and understanding of healthcare informatics
US7769751B1 (en) * 2006-01-17 2010-08-03 Google Inc. Method and apparatus for classifying documents based on user inputs
US7571162B2 (en) 2006-03-01 2009-08-04 Microsoft Corporation Comparative web search
US7809665B2 (en) * 2006-03-22 2010-10-05 Sony Corporation Method and system for transitioning from a case-based classifier system to a rule-based classifier system
US20080256026A1 (en) 2006-10-17 2008-10-16 Michael Glen Hays Method For Optimizing And Executing A Query Using Ontological Metadata
US7827172B2 (en) 2007-03-14 2010-11-02 Yahoo! Inc. “Query-log match” relevance features
US20080235209A1 (en) * 2007-03-20 2008-09-25 Samsung Electronics Co., Ltd. Method and apparatus for search result snippet analysis for query expansion and result filtering

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6144964A (en) * 1998-01-22 2000-11-07 Microsoft Corporation Methods and apparatus for tuning a match between entities having attributes
US6345264B1 (en) * 1998-01-22 2002-02-05 Microsoft Corporation Methods and apparatus, using expansion attributes having default, values, for matching entities and predicting an attribute of an entity
US6018738A (en) * 1998-01-22 2000-01-25 Microsft Corporation Methods and apparatus for matching entities and for predicting an attribute of an entity based on an attribute frequency value
US20070038610A1 (en) * 2001-06-22 2007-02-15 Nosa Omoigui System and method for knowledge retrieval, management, delivery and presentation
US20050149538A1 (en) * 2003-11-20 2005-07-07 Sadanand Singh Systems and methods for creating and publishing relational data bases
US20050154713A1 (en) * 2004-01-14 2005-07-14 Nec Laboratories America, Inc. Systems and methods for determining document relationship and automatic query expansion
US20050234881A1 (en) * 2004-04-16 2005-10-20 Anna Burago Search wizard
US20050289100A1 (en) * 2004-06-25 2005-12-29 International Business Machines Corporation Techniques for representing relationships between queries
US20060242130A1 (en) * 2005-04-23 2006-10-26 Clenova, Llc Information retrieval using conjunctive search and link discovery
US20080140616A1 (en) * 2005-09-21 2008-06-12 Nicolas Encina Document processing
US7792858B2 (en) * 2005-12-21 2010-09-07 Ebay Inc. Computer-implemented method and system for combining keywords into logical clusters that share similar behavior with respect to a considered dimension
US20080189269A1 (en) * 2006-11-07 2008-08-07 Fast Search & Transfer Asa Relevance-weighted navigation in information access, search and retrieval
US7966305B2 (en) * 2006-11-07 2011-06-21 Microsoft International Holdings B.V. Relevance-weighted navigation in information access, search and retrieval
US7933904B2 (en) * 2007-04-10 2011-04-26 Nelson Cliff File search engine and computerized method of tagging files with vectors
US20080306914A1 (en) * 2007-06-05 2008-12-11 Search Capital Ltd Method and system for performing a search

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140136504A1 (en) * 2012-11-12 2014-05-15 Microsoft Corporation Multidimensional search architecture

Also Published As

Publication number Publication date Type
US8639682B2 (en) 2014-01-28 grant
CA2688642A1 (en) 2010-06-29 application
US20100169375A1 (en) 2010-07-01 application

Similar Documents

Publication Publication Date Title
Cody et al. The integration of business intelligence and knowledge management
US7548936B2 (en) Systems and methods to present web image search results for effective image browsing
US7647306B2 (en) Using community annotations as anchortext
US20120117059A1 (en) Ranking Authors in Social Media Systems
US20040215663A1 (en) Media agent
US20080114644A1 (en) Convergence Of Terms Within A Collaborative Tagging Environment
US20060282411A1 (en) System and method for performing a high-level multi-dimensional query on a multi-structural database
US20080222105A1 (en) Entity recommendation system using restricted information tagged to selected entities
US20050165777A1 (en) System and method for a unified and blended search
US20110016121A1 (en) Activity Based Users' Interests Modeling for Determining Content Relevance
US20090012969A1 (en) Systems and processes for evaluating webpages
US20070055680A1 (en) Method and system for creating a taxonomy from business-oriented metadata content
US20130311485A1 (en) Method and system relating to sentiment analysis of electronic content
US20110218946A1 (en) Presenting content items using topical relevance and trending popularity
US20070179945A1 (en) Determining relevance of electronic content
US20060122978A1 (en) Entity-specific tuned searching
US20100114899A1 (en) Method and system for business intelligence analytics on unstructured data
US20100153371A1 (en) Method and apparatus for blending search results
Blumberg et al. The problem with unstructured data
US20120150874A1 (en) Methods and apparatus for providing information of interest to one or more users
USRE41899E1 (en) System for ranking the relevance of information objects accessed by computer users
US20100211924A1 (en) Discovering and exploiting relationships in software repositories
US20110179025A1 (en) Social and contextual searching for enterprise business applications
US8316012B2 (en) Apparatus and method for facilitating continuous querying of multi-dimensional data streams
US20130191723A1 (en) Web Browser Device for Structured Data Extraction and Sharing via a Social Network