Abstract | Some languages lack large knowledge bases and good discriminative features for Name Entity Recognition ( NER ) that can generalize to previously unseen named entities. |
Introduction | Named Entity Recognition ( NER ) is essential for a variety of Natural Language Processing (NLP) applications such as information extraction. |
Introduction | There has been a fair amount of work on NER for a variety of languages including Arabic. |
Introduction | To train an NER system, some of the following feature types are typically used (Benajiba and Rosso, 2008; Nadeau and Sekine, 2009): |
Abstract | Translated bi-texts contain complementary language cues, and previous work on Named Entity Recognition ( NER ) has demonstrated improvements in performance over monolingual taggers by promoting agreement of tagging decisions between the two languages. |
Abstract | We observe that NER label information can be used to correct alignment mistakes, and present a graphical model that performs bilingual NER tagging jointly with word alignment, by combining two monolingual tagging models with two unidirectional alignment models. |
Abstract | We design a dual decomposition inference algorithm to perform joint decoding over the combined alignment and NER output space. |
Bilingual NER by Agreement | We assume access to two monolingual linear-chain CRF-based NER models that are already trained. |
Introduction | We study the problem of Named Entity Recognition ( NER ) in a bilingual context, where the goal is to annotate parallel bi-texts with named entity tags. |
Introduction | (2012) have also demonstrated that bi-texts annotated with NER tags can provide useful additional training sources for improving the performance of standalone monolingual taggers. |
Introduction | In this work, we first develop a bilingual NER model (denoted as BI-NER) by embedding two monolingual CRF-based NER models into a larger undirected graphical model, and introduce additional edge factors based on word alignment (WA). |
Background | Compared to shallow (POS, NER ) structured retrieval, deep structures need more processing power and smoothing, but might also be more precise. |
Introduction | This will be our off-the-shelf QA system, which recognizes the association between question type and expected answer types through various features based on e.g., part-of-speech tagging (POS) and named entity recognition ( NER ). |
Introduction | For instance, line 2 in Table 1 says that if there is a when question, and the current token’s NER label is DATE, then it is likely that this token is tagged as ANS. |
Introduction | IR can easily make use of this knowledge: for a when question, IR retrieves sentences with tokens labeled as DATE by NER , or POS tagged as CD. |
Method | We let the trained QA system guide the query formulation when performing coupled retrieval with Indri (Strohman et al., 2005), given a corpus already annotated with POS tags and NER labels. |
Method | For instance, the NER tagger we used divides location into two categories: GPE (geo locations) and LOC |
Method | Take the previous where question, besides NER[O]=GPE and NER[O]=LOC, we also found oddly N ER[O]=PERSON an important feature, due to that the NER tool sometimes mistakes PERSON for LOC. |
Data | Table 1 gives statistics for both corpora, including the number and ambiguity of gold standard toponyms for both as well as NER identified to- |
Data | ponyms for TR-CONLL.4 We use the pre-trained English NER from the OpenNLP project.5 |
Evaluation | False positives occur when the NER incorrectly predicts a toponym, and false negatives occur when it fails to predict a toponym identified by the annotator. |
Results | In this case, the ORACLE results are less than 100% due to the limitations of the NER, and represent the best possible results given the NER we used. |
Results | Results on TR-CONLL indicate much higher performance than the resolvers presented by Leidner (2008), whose F-scores do not exceed 36.5% with either gold or NER toponyms.7 TRC-TEST is a subset of the documents Leidner uses (he did not split development and test data), but the results still come from overlapping data. |
Results | However, our evaluation is more penalized since SPIDER loses precision for NER’s false positives (Jack London as a location) while Leidner only evaluated on actual locations. |
Toponym Resolvers | Given a set of toponyms provided via annotations or identified using NER , a resolver must select a candidate location for each toponym (or, in some cases, a resolver may abstain). |
Toponym Resolvers | 4States and countries are not annotated in CWAR, so we do not evaluate end-to-end using NER plus toponym resolution for it as there are many (falsely) false positives. |
Abstract | To evaluate our method, we use the word clusters in an NER system and demonstrate a statistically significant improvement in F1 score when using bilingual word clusters instead of monolingual clusters. |
Conclusions | We have shown that improvement in clustering can be obtained across a range of language pairs, evaluated in terms of their value as features in an extrinsic NER task. |
Experiments | Our evaluation task is the German corpus with NER annotation that was created for the shared task at CoNLL-2003 3. |
Experiments | Table 1 shows the performance of NER when the word clusters are obtained using only the bilingual information for different language pairs. |
Experiments | We varied the weight of the bilingual objective (/3) from 0.05 to 0.9 and observed the effect in NER performance on English-German language pair. |
RSP: A Random Walk Model for SP | Random confounder (RND) most closes to the realistic case; While nearest confounder ( NER ) is reproducible and it avoids frequency bias (Chambers and Jurafsky, 2010). |
RSP: A Random Walk Model for SP | In this work, we employ both RND and NER confounders: 1) for RND, we randomly select |
RSP: A Random Walk Model for SP | 2) for NER , firstly we sort the arguments by their frequency. |
Experiments | We use the standard automatic parses and NER tags for each document. |
Introduction | We evaluate our system on the dataset from the CoNLL 2011 shared task using three different types of properties: synthetic oracle properties, entity phi features (number, gender, animacy, and NER type), and properties derived from unsupervised clusters targeting semantic type information. |
Models | Agreement features: Gender, number, animacy, and NER type of the current mention and the antecedent (separately and conjoined). |
Models | Each mention 2' has been augmented with a single property node pi E {1, ..., The unary B factors encode prior knowledge about the setting of each pi; these factors may be hard (I will not refer to a plural entity), soft (such as a distribution over named entity types output by an NER tagger), or practically uniform (e. g. the last name Smith does not specify a particular gender). |
Creating Text-to-text Relations via Twitter/News Features | Directly applying Named Entity Recognition ( NER ) tools on news titles or |
Creating Text-to-text Relations via Twitter/News Features | Accordingly, we first apply the NER tool on news summaries, then label named entities in the tweets in the same way as labeling the hashtags: if there is a string in the tweet that matches a named entity from the summaries, then it is labeled as a named entity in the tweet. |
Creating Text-to-text Relations via Twitter/News Features | The noise introduced during automatic NER accumulates much faster given the large number of named entities in news data. |
Introduction | Named entities acquired from a news document, typically with high accuracy using Named Entity Recognition [ NER ] tools, may be particularly informative. |
Distant Supervision | Generally, the quality of NER is crucial in this task. |
Features | As standard named entity recognition ( NER ) systems do not capture categories that are relevant to the movie domain, we opt for a lexicon-based approach similar to (Zhuang et al., 2006). |
Features | Many entries are unsuitable for NER , e.g., dog is frequently listed as a character. |
Features | This rule has precedence over NER, so if a name matches a labeled entity, we do not attempt to label it through NER . |