Skip to content

Latest commit

 

History

History
130 lines (87 loc) · 7.08 KB

README.md

File metadata and controls

130 lines (87 loc) · 7.08 KB
Raw

IntentsKB: A Knowledge Base of Entity-Oriented Search Intents

This repository provides resources developed within the following article:

D. Garigliotti and K. Balog. IntentsKB: A Knowledge Base of Entity-Oriented Search Intents. In: Proceedings of the 27th ACM International Conference on Information and Knowledge Management (CIKM '18). ACM. Turin, Italy. October 2018. DOI: 10.1145/3269206.3269257

You can get the author version of the article here.

Abstract

We address the problem of constructing a knowledge base of entity-oriented search intents. Search intents are defined on the level of entity types, each comprising of a high-level intent category (property, website, service, or other), along with a cluster of query terms used to express that intent. These machine-readable statements can be leveraged in various applications, e.g., for generating entity cards or query recommendations. By structuring service-oriented search intents, we take one step towards making entities actionable. The main contribution of this paper is a pipeline of components we develop to construct a knowledge base of entity intents. We evaluate performance both component-wise and end-to-end, and demonstrate that our approach is able to generate high-quality data.

1. Refiner Acquisition

Data obtained in the refiners acquisition pipeline stage (Sect. 5.1 of the paper) are placed under the directory data/1-refiners_acquisition/.

  • A tab-separated file 1-top_entities_per_type/<type>/entity_pageviews.tsv contains the top 1,000- most popular entities of the Freebase type <type> . Each row in the file is in the format:
<entity>	<pageviews>
  • After decompressing 2-API_query_suggestions.tar.gz (into a directory of 1.3 GB of size), a JSON file 2-API_query_suggestions/<type>/<i>.json contains the API query suggestions obtained for the (i+1)-th entity in the entities file of the type <type> (i >= 0). The content is of the shape:
{
    "<i>": <list of API query suggestions>
}
  • A tab-separated file 3-type_level_query_patterns/<type>.tsv contains the type-level query patterns for type <type>, decreasingly sorted by frequency (total number of API query suggestions for entities in <type>). Each row in the file is in the format:
<type-level query pattern>	<frequency>

2. Refiner Categorization

Intent categories predicted in the refiners categorization pipeline stage (Sect. 5.2 of the paper) are contained in the tab-separated file data/2-refiners_categorization/predictions.tsv. Each row in the file is of the shape:

<type>	<refiner>	<category>

Search results

Additionally, the results of searching for entity-bearing queries with Google Custom Search Engine (CSE) (Sect. 4.2 of the paper) are contained under 2 compressed directories (freq_{5_to_9,10+}.tar.gz), shared in this link. Once decompressed:

  • A file cse_results-freq_10+/<type>/query_<i>.json contains the JSON response after searching for the actual entity-bearing query that results from instantiating the (i+1)-th type-level query pattern with the most prominent entity of that type. (Such a pattern is the one in the (i+1)-th row of data/1-refiners_acquisition/3-type_level_query_patterns/<type>.tsv.)
  • Analogously, a file cse_results-freq_5_to_9/<type>/query_<i>.json contains the JSON responses for less frequent type-level refiners. Specifically, it contains the response for the (i+1)-th type-level query pattern, considering only the patterns with frequency in the range [5..9] (i.e., the pattern in the (i + N_<type> + 1)-th row of data/1-refiners_acquisition/3-type_level_query_patterns/<type>.tsv, with N_<type> being the number of query patterns for <type> with frequency larger than 9).

Each JSON file is in this format:

{
    "1_type": <type>,
    "2_tl_q": <type-level query pattern>,
    "3_ae_q": <actual entity-bearing query>,
    "4_f": <frequency of type-level query pattern>,
    "5_results": {
        "items": <list of top-10 search results>,
        ...
    }
}

Table of lexical features

This is the full list of lexical features for refiner categorization (Sect. 4.2.1 of the paper):

Id Feature description Value
1 Length of the top result title, in characters {1, ..., }
2 Avg. length of the 10 result titles, in characters {1, ..., }
3 Length of the top result snippet, in characters {1, ..., }
4 Avg. length of the 10 result snippets, in characters {1, ..., }
5 Number of / in the top result URL {0, ..., }
6 Avg. number of / in the top 10 result URLs {0, ..., }
7 Size of the URL domains set among all top 10 result URLs. {1, ..., 10}
8 Jaro distance between the refiner and the top result URL [0..1]
9 Avg. Jaro distance between the refiner and each of the top 10 result URLs [0..1]
10 Avg. Jaro distance between all the top 10 result URLs [0..1]
11-13 {Max., Min., Mean} aggregated term-to-term Jaro distance between the refiner and the top result title [0..1]
14-16 Avg. {max., min., mean} aggregated term-to-term Jaro distance between the refiner and the top 10 result titles [0..1]

3. Intent Discovery

Entity-oriented search intents discovered in the intents discovery pipeline stage (Sect. 5.3 of the paper) are stored in the tab-separated file data/3-intents_discovery/clustering.tsv. Each row in the file is in the format:

<type>	<category>	<cluster ID>	<cluster label>	<refiner>	<similarity>

4. IntentsKB Knowledge Base

The final knowledge base (Sect. 5.4 of the paper) is stored in the tab-separated file intentsKB/quadruples.tsv, which contains rows each of the shape as follows (Sect. 3 of the paper):

<intent ID>	<predicate>	<object>	<confidence>

Here, <object> is - for the predicate isAProfile (each meaning a quadruple that only states the confidence of the intent profile; cf. Eq. (1) of the paper).

Citation

If you use the resources presented in this repository, please cite:

@inproceedings{Garigliotti:2018:IKB,
 author =    {Dar\'{i}o Garigliotti and Krisztian Balog},
 title =     {IntentsKB: A Knowledge Base of Entity-Oriented Search Intents},
 booktitle = {Proceedings of the 27th ACM International Conference on Information and Knowledge Management},
 series =    {CIKM '18},
 year =      {2018},
 pages =     {1659--1662},
 doi =       {10.1145/3269206.3269257},
 publisher = {ACM}
}

Contact

Should you have any questions, please contact Darío Garigliotti at dario.garigliotti[AT]uis.no (with [AT] replaced by @).