Patent Application
20120023119
The present invention relates to a system for searching data and in particular for a system that can analyse a query and match the query contents to one or more relevant multimedia resources.
With the advent of the world wide web, the able to find and retrieve useful and relevant information has become more important and nowadays search engines are essential tools to find information. Search engines such as Google or Yahoo work by virtue of “mass availability”, i.e. the content is there in abundance but the most difficult task is to index as much of the content as possible. Content can retrieved by providing short descriptions for example, one or two keywords, to a search engine. Similarly, search engines for the enterprise became essential because corporate intranets began to suffer from information overload. Popular intranet search engines include Autonomy, Fast, Endeca, etc. which typically work with concept association algorithms, where keywords are associated to larger concepts and content is easily browsed and refined to be able to reach information.
The use of online multimedia content has been popularised by Youtube, which allows users to search for video clips of interest. Such searches require, like text searches, short queries, for example, one to three keywords, which can be matched to keywords that are associated with multimedia content. Similar searching techniques can be applied to material such as podcasts and photos stored in online services such as Flickr.
According to a first aspect of the present invention there is provided a method of searching data using an augmented user query, the method comprising the steps of: a) augmenting a user query, wherein the user query augmentation comprises the steps of: i) receiving a query from a user; ii) analysing the content of the user query received in step i); iii) presenting one or more selection options to the user based on the results of the analysis performed in step ii); iv) receiving the one or more selections from the user; and v) augmenting the query in accordance with the selections received in step iv); b) searching a data index for terms that match with the augmented query, the data index comprising multimedia data, the multimedia data being associated with one or more metadata tags; and c) returning to the user one or more search results which match the terms of the augmented query.
In step ii) of the query augmentation, the user query content may be analysed grammatically; the grammatical analysis may determine the significant nouns and verbs which comprise the user query content. Also, in step ii) of the query augmentation, the user query content may be analysed for the presence of specific entities. Furthermore, in step ii) of the query augmentation the user query content may be analysed contextually. In step iii) of the query augmentation, the user may be presented with one or more of the following as a selection option: a product category, a product name or an action. The metadata tags can be assigned by users to the associated multimedia data.
According to a second aspect of the present invention there is provided a tangible computer program product comprising computer executable code for performing a method as described above.
According to a third aspect of the present invention there is provided a system for searching data, the system configured to execute a method according to a method as described above.
Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings in which:
a shows a natural language query which is provided to the online forum;
b shows a graphical depiction of the further keywords that may be selected by the user; and
The content repository also comprises a multimedia content store 36 which holds user generated multimedia content which has been supplied by members of a community of active users 20. This community of active users may also provide further data which relates to the uploaded user generated multimedia content; this may take the form of applying ‘tags’ which provide a descriptor of the multimedia content, ratings or recommendations for the multimedia content, links to other online content which may relate to the multimedia content, etc. The multimedia content store 36 may comprise a multimedia search engine 38 which enables a user to search the multimedia content data held in the multimedia content store 36 and any data relating to the multimedia content data. It will be understood that members of the community of active users may post queries and/or responses to the forum 32 and use the search engine 34 to search the content held in the forum.
The present invention provides a system 100 which is able to interpret queries that are submitted to the forum 32 and then provide an answer to the query, for example by directing the user which submitted the query to the most relevant item of multimedia content that is held in the multimedia content store 36.
Referring to
The operation of a system and a method according to the present invention will now be described in more detail with reference to
In operation, a user will enter a query, for example by posting a query on the forum 32. This query will firstly be sent to the query analysis element 110 to determine the contents of the query. For example, the verbs and nouns in the query may be recognised, along with a determination of whether they are strong verbs or nouns. The query analysis element may also determine whether the query comprises any product or brand names; in further alternatives the query analysis element may identify other entities that are comprised within the query such as, for example, addresses, contact details, product characteristics (such as price, internet download speed) etc. It will be understood that this, list is not exhaustive and that further entities that may be of interest may be identified within a query.
An example of the function of the query analysis element will now be given. If a user submits the query “I can not get my BT Hub Phone 1010 to register to the BT Hub” then this query can be parsed to give the following structure:
Such a query representation can be obtained through the application of a part of speech tagger (such as, for example the Brill tagger [which can be obtained from http://www.cs.jhu.edu/˜brill/RBT1—14.tar.Z]), which returns a grammatical analysis of the query. The query can be broken down with each word being associated with a particular speech component, such as, for example: noun, verb, adjective, preposition, pronoun, adverb or conjunction. These results can be the subject of further grammatical analysis such that, in turn, identifies punctuation, verb compositions (e.g. negative, affirmative, phrasal verbs), nouns (phrases, compounds, proper nouns) and adjectives.
It will be understood by the person skilled in the art that other part of speech taggers may be used without effecting the operation of the present invention and that the part of speech tagger must be specific to the language of the query. It will also be understood that if languages other than English are to be analysed then it may be necessary to make some adjustments to the query analysis engine (e.g. phrasal verbs may not exist in other languages, etc).
Once the grammatical analysis is complete, then the query analysis element 110 can undertake entity discovery on the query. Depending on their type, some entities can be discovered with direct matching (for example, a product name can be easily identified) however other entities such as an address may require a soft pattern matching technique. The soft matching process involves a training set of addresses with a finely defined grammar (street no, street name, post code, etc). This grammar can then be matched systematically in the document to highlight a full or an partial match—successfully discovered addresses can then be marked up and displayed to a user. For example, the bit rate and the price of broadband products will require soft matching, whereas product matching would require a hard match (although it may be possible to provide some flexibility to allow for the mis-spelling of product names, this would need to be pre-defined through the provision of additional search terms).
Analysis of this representation of the query gives the following results:
Once the contents of a query have been analysed, it is possible to form a computer representation of that query. It has been observed that sometimes it is difficult to answer queries in a satisfactory manner because a query may be phrased in a different context to that which is used by a suitable answer. For example, a query may use one set of keywords or terminology to describe the problem whereas an answer may use different keywords when describing the same problem.
The results obtained from the query analysis element 110 may be processed by the query representation element 120. The query analysis element has identified the most important words or attributes of the query and ranked them in order of significance. The query representation element is then able to extract the keywords that are of most significance and can then present the user with a selection of potential alternative or additional keywords that may be of use.
These keywords that are presented to a user are keywords that have been detected as strong keywords within the original query (or keywords that share the same root as a keyword found in the original query), keywords that are found in other queries or postings, or keywords that are in found in the online forum and that can be searched by the forum search engine. It is possible, using a statistical analysis, to calculate a link between strong keywords and other keywords (or phrases) for a given context such that if an association exceeds a threshold then, the keywords (or phrases) may be considered a good candidate. For details on the calculation of the threshold see G Ducatel et al, “Hybrid Ontology and Keyword Matching Indexing System” from the Proceedings of WWW2006, published May 2006, which can be found at http://www-sop.inria.fr/acacia/WORKSHOPS/IntraWebs2006/Ducatel_Intrawebs2006 pdf. It will be understood that the quality of keyword suggestion will increase as the size of the text repository increases. It has been found that a minimum of 2 MB of user generated data is required to provide suggestions of a consistent quality.
When a query is posted to the forum it is likely to receive a number of answers or responses from other members of the community of active users. As answers are posted, it is possible to augment the query representation by incorporating into the query representation keywords that are frequently posted in answers to the query. The query augmentation element 130 stores the query representation and the additional keywords that are added to the representation. Users may confirm that a keyword is relevant to a query when they posts an answer or alternatively conventional keyword frequency analysis and word weighting techniques can be used to extract keywords frequently associated with a particular query.
The augmented query can be used to query the multimedia content store 36, for example by sending a request to the multimedia search engine 38. The content held in the multimedia content store will normally have one or more tags associated with it, for example by the producers of the content and/or by viewers or users, such that the search engine is able to match queries to the content.
For example, all multimedia resources for which the product category, the product name, and the process match the augmented query can be ‘pushed’ into the appropriate part of the forum, for example by providing a link to the multimedia resources next to the query. The resources will be visible to all users browsing the forum and the responses of users can be used to rank the multimedia resources, for example by ordering the resources in accordance with the number of times that the resource has been accessed or in accordance with rankings that have been supplied by users.
It will be understood from the foregoing discussion that the present invention will be implemented in software using a conventional client-server computer architecture. The functionality that enables a general purpose computing device (or devices) to perform the method of the present invention may be provided by providing one or more additional computer programs or applications. It will be understood that such software may be deployed to such a computing device via download, for example via the internet, or on some physical media, for example, DVD, CD-ROM, USB memory stick, etc.
| Number | Date | Country | Kind |
|---|---|---|---|
| 092509454 | Mar 2009 | EP | regional |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/GB2010/000582 | 3/26/2010 | WO | 00 | 9/30/2011 |
