Identifying and Exploiting Target Entity Type Information for Ad Hoc Entity Retrieval
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
code
evaluation
tti-dbpedia-ltr_features
README.md

README.md

Identifying and Exploiting Target Entity Type Information for Ad Hoc Entity Retrieval

This repository provides resources developed within the following article:

D. Garigliotti, F. Hasibi, and K. Balog. Identifying and Exploiting Target Entity Type Information for Ad Hoc Entity Retrieval. Information Retrieval Journal, (), 1-39. DOI: 10.1007/s10791-018-9346-x

Abstract

Today, the practice of returning entities from a knowledge base in response to search queries has become widespread. One of the distinctive characteristics of entities is that they are typed, i.e., assigned to some hierarchically organized type system (type taxonomy). The primary objective of this paper is to gain a better understanding of how entity type information can be utilized in entity retrieval. We perform this investigation in two settings: firstly, in an idealized ``oracle'' setting, assuming that we know the distribution of target types of the relevant entities for a given query; and secondly, in a realistic scenario, where target entity types are identified automatically based on the keyword query. We perform a thorough analysis of three main aspects: (i) the choice of type taxonomy, (ii) the representation of hierarchical type information, and (iii) the combination of type-based and term-based similarity in the retrieval model. Using a standard entity search test collection based on DBpedia, we show that type information can significantly and substantially improve retrieval performance, \response{yielding up to 67% relative improvement in terms of NDCG@10 over a strong text-only baseline in an oracle setting. We further show that using automatic target type detection, we can outperform the text-only baseline by 44% in terms of NDCG@10. This is as good as, and sometimes even better than, what is attainable by using explicit target type information provided by humans. These results indicate that identifying target entity types of queries is challenging even for humans and attests to the effectiveness of our proposed automatic approach.

Repository Files

This repository is structured as follows:

  • evaluation/: all data needed to reproduce the evaluation results presented in the article.

  • code/type_aware_ranking.py: a Python file with the code needed for obtaining an entity ranking, from a term-based baseline and entity type information, for a particular retrieval configuration.

  • tti-dbpedia-ltr_features/features.json: a JSON file containing the training set with all precomputed features for our Learning-to-Rank Target Type Identification approach.

Entity Retrieval results

Entity Retrieval results presented in the article can be obtained as explained below.

Make sure you are in entity_retrieval/ subdirectory:

$ pwd
/path/to/repo/irj-types/evaluation/entity_retrieval

Table 6

These results correspond to evaluate an entity retrieval runfile under the (uncompressed) directory runs/type_aware/tables_6_and_7/ versus the default qrels data/qrels/qrels.txt:

$ /path/to/trec_eval -c -m ndcg_cut.10,100 data/qrels/qrels.txt runs/baseline/sdm.txt  # for the baseline
$ /path/to/trec_eval -c -m ndcg_cut.10,100 data/qrels/qrels.txt runs/type_aware/tables_6_and_7/<model>/<representation>/<taxonomy>[_<interpolation_lambda>].txt  # for type-aware

Table 7

Similarly to Table 6, but here a runfile is evaluated against the taxonomy-specific qrels data/qrels/qrels_<taxonomy>.txt:

$ /path/to/trec_eval -c -m ndcg_cut.10,100 data/qrels/qrels_<taxonomy>.txt runs/baseline/sdm.txt  # for the baseline
$ /path/to/trec_eval -c -m ndcg_cut.10,100 data/qrels/qrels_<taxonomy>.txt runs/type_aware/tables_6_and_7/<model>/<representation>/<taxonomy>[_<interpolation_lambda>].txt  # for type-aware

Table 12

These results correspond to evaluate an entity retrieval runfile under the (uncompressed) directory runs/type_aware/table_12/ versus the DBpedia-specific qrels data/qrels/qrels_dbpedia.txt.

  • The subdirectory <TTI>/ corresponds to the Target Type Identification (TTI) model used for automatically identifying target types. <TTI> can be: ec/bm25/k_10, tc/lm, or ltr.
$ /path/to/trec_eval -c -m ndcg_cut.10,100 data/qrels/qrels_dbpedia.txt runs/baseline/sdm.txt  # for the baseline
$ /path/to/trec_eval -c -m ndcg_cut.10,100 data/qrels/qrels_dbpedia.txt runs/type_aware/table_12/auto_types/<TTI>/<model>/<representation>/dbpedia[_<interpolation_lambda>|_<strict_filtering_k>].txt  # for automatically identified types
$ /path/to/trec_eval -c -m ndcg_cut.10,100 data/qrels/qrels_dbpedia.txt runs/type_aware/table_12/oracle2_types/<model>/<representation>/dbpedia[_<interpolation_lambda>].txt  # for Oracle 2
$ /path/to/trec_eval -c -m ndcg_cut.10,100 data/qrels/qrels_dbpedia.txt runs/type_aware/table_12/oracle1_types/<model>/<representation>/dbpedia[_<interpolation_lambda>].txt  # for Oracle 1

Target Type Identification results

Target Type Identification (TTI) results presented in Table 7 can be obtained as follows.

Make sure you are now in tti/ subdirectory and run trec_eval for a TTI runfile under runs-table_10/ versus the DBpedia type ranking qrels (from SIGIR'17 test collection):

$ pwd
/path/to/repo/irj-types/evaluation/tti
$ /path/to/trec_eval -c -m ndcg_cut.1,5 data/qrels/qrels-tti-dbpedia.txt runs-table_10/<TTI_model>.txt

Contact

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