Joint Inference of Named Entity Recognition and Normalization for Tweets
Liu, Xiaohua and Zhou, Ming and Zhou, Xiangyang and Fu, Zhongyang and Wei, Furu

Article Structure

Abstract

Tweets represent a critical source of fresh information, in which named entities occur frequently with rich variations.

Introduction

Tweets, short messages of less than 140 characters shared through the Twitter service 1, have become an important source of fresh information.

Related Work

Related work can be divided into two categories: NER and NEN.

Task Definition

A tweet is a short text message with no more than 140 characters.

Our Method

In contrast to existing work, our method jointly conducts NER and NEN for multiple tweets.

Experiments

We manually annotate a data set to evaluate our method.

Conclusions and Future work

We study the task of NEN for tweets, a new genre of texts that are short and prone to noise.

Topics

NER

Appears in 40 sentences as: NER (44)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. Two main challenges are the errors propagated from named entity recognition ( NER ) and the dearth of information in a single tweet.
    Page 1, “Abstract”
  2. We evaluate our method on a manually annotated data set, and show that our method outperforms the baseline that handles these two tasks separately, boosting the F1 from 80.2% to 83.6% for NER , and the Accuracy from 79.4% to 82.6% for NEN, respectively.
    Page 1, “Abstract”
  3. As a result, the task of named entity recognition ( NER ) for tweets, which aims to identify mentions of rigid designators from tweets belonging to named-entity types such as persons, organizations and locations (2007), has attracted increasing research interest.
    Page 1, “Introduction”
  4. Traditionally, NEN is regarded as a septated task, which takes the output of NER as its input (Li et al., 2002; Cohen, 2005; Jijkoun et al., 2008; Dai et al., 2011).
    Page 1, “Introduction”
  5. One limitation of this cascaded approach is that errors propagate from NER to NEN and there is no feedback from NEN to NER .
    Page 1, “Introduction”
  6. We propose jointly conducting NER and NEN on multiple tweets using a graphical model, to address these challenges.
    Page 2, “Introduction”
  7. Intuitively, improving the performance of NER boosts the performance of NEN.
    Page 2, “Introduction”
  8. On the other hand, NEN can guide NER .
    Page 2, “Introduction”
  9. This will help NER to assign them different types of labels.
    Page 2, “Introduction”
  10. Figure 1: A factor graph that jointly conducts NER and NEN on multiple tweets.
    Page 2, “Introduction”
  11. It is these normalization variables that enable us to conduct NER and NEN jointly.
    Page 2, “Introduction”

See all papers in Proc. ACL 2012 that mention NER.

See all papers in Proc. ACL that mention NER.

Back to top.

named entities

Appears in 14 sentences as: named entities (9) named entity (7)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. Tweets represent a critical source of fresh information, in which named entities occur frequently with rich variations.
    Page 1, “Abstract”
  2. We study the problem of named entity normalization (N EN ) for tweets.
    Page 1, “Abstract”
  3. Two main challenges are the errors propagated from named entity recognition (NER) and the dearth of information in a single tweet.
    Page 1, “Abstract”
  4. As a result, the task of named entity recognition (NER) for tweets, which aims to identify mentions of rigid designators from tweets belonging to named-entity types such as persons, organizations and locations (2007), has attracted increasing research interest.
    Page 1, “Introduction”
  5. entities and then uses a distantly supervised approach based on LabeledLDA to classify named entities .
    Page 1, “Introduction”
  6. However, named entity normalization (NEN) for tweets, which transforms named entities mentioned in tweets to their unambiguous canonical forms, has not been well studied.
    Page 1, “Introduction”
  7. Owing to the informal nature of tweets, there are rich variations of named entities in them.
    Page 1, “Introduction”
  8. (2011), every named entity in tweets has an average of 3.3 variations 2.
    Page 1, “Introduction”
  9. (2010) use Amazons Mechanical Turk service 3 and CrowdFlower 4 to annotate named entities in tweets and train a CRF model to evaluate the effectiveness of human labeling.
    Page 3, “Related Work”
  10. (2011) rebuild the NLP pipeline for tweets beginning with POS tagging, through chunking, to NER, which first exploits a CRF model to segment named entities and then uses a distantly supervised approach based on LabeledLDA to classify named entities .
    Page 3, “Related Work”
  11. Unlike this work, our work detects the boundary and type of a named entity simultaneously using sequential labeling techniques.
    Page 3, “Related Work”

See all papers in Proc. ACL 2012 that mention named entities.

See all papers in Proc. ACL that mention named entities.

Back to top.

entity mention

Appears in 5 sentences as: entities mentioned (1) entity mention (2) entity mentions (2)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. However, named entity normalization (NEN) for tweets, which transforms named entities mentioned in tweets to their unambiguous canonical forms, has not been well studied.
    Page 1, “Introduction”
  2. Given a set of tweets, e. g., tweets within some period or related to some query, our task is: 1) To recognize each mention of entities of predefined types for each tweet; and 2) to restore each entity mention into its unambiguous canonical form.
    Page 4, “Task Definition”
  3. Given each pair of entity mentions , decide whether they denote the same entity.
    Page 4, “Task Definition”
  4. For any entity mention , two annotators independently annotate its canonical form.
    Page 7, “Experiments”
  5. This explains the cases where our system correctly links multiple entity mentions but fails to generate canonical forms.
    Page 8, “Experiments”

See all papers in Proc. ACL 2012 that mention entity mention.

See all papers in Proc. ACL that mention entity mention.

Back to top.

Feature Set

Appears in 5 sentences as: Feature Set (2) feature set (1) feature sets (2)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. (2) {$21521 and {$225521 are two feature sets .
    Page 5, “Our Method”
  2. 4.3.1 Feature Set One: {$21) 5:11
    Page 6, “Our Method”
  3. 4.3.2 Feature Set Two: {$22) 5:21
    Page 6, “Our Method”
  4. Table 4: Overall F1 (%) of NER and Accuracy (%) of N EN with different feature sets .
    Page 8, “Experiments”
  5. Table 4 shows the overall performance of our method with various feature set combinations, where F0, F; and F9 denote the orthographic features, the lexical features, and the gazetteer-related features, respectively.
    Page 8, “Experiments”

See all papers in Proc. ACL 2012 that mention Feature Set.

See all papers in Proc. ACL that mention Feature Set.

Back to top.

CRF

Appears in 4 sentences as: CRF (4)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. (2011) develop a system that exploits a CRF model to segment named
    Page 1, “Introduction”
  2. A linear CRF model
    Page 3, “Related Work”
  3. (2010) use Amazons Mechanical Turk service 3 and CrowdFlower 4 to annotate named entities in tweets and train a CRF model to evaluate the effectiveness of human labeling.
    Page 3, “Related Work”
  4. (2011) rebuild the NLP pipeline for tweets beginning with POS tagging, through chunking, to NER, which first exploits a CRF model to segment named entities and then uses a distantly supervised approach based on LabeledLDA to classify named entities.
    Page 3, “Related Work”

See all papers in Proc. ACL 2012 that mention CRF.

See all papers in Proc. ACL that mention CRF.

Back to top.

entity type

Appears in 4 sentences as: entity type (3) entity types (2)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. To resolve NER, we assign a label to each word in a tweet, indicating both the boundary and entity type .
    Page 4, “Our Method”
  2. tiflfil, respectively, refer to the same entity if and only if: 1) The two mentions share the same entity type ; 2) is a substring of film” or vise versa; ands) 23%.. = 1, z' = .27. andj = .71.--- .32, if 277%” exists.
    Page 5, “Our Method”
  3. Table 3 reports the NER performance of our method for each entity type, from which we see that our system consistently yields better F1 on all entity types than S B R. We also see that our system boosts the F1 for ORGANIZATION most significantly, reflecting the fact that a large number of organizations that are incorrectly labeled as PERSON by S B R, are now correctly recognized by our method.
    Page 7, “Experiments”
  4. Table 3: F1 (%) of NER on different entity types .
    Page 8, “Experiments”

See all papers in Proc. ACL 2012 that mention entity type.

See all papers in Proc. ACL that mention entity type.

Back to top.

manually annotated

Appears in 4 sentences as: manually annotate (1) manually annotated (3)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. We evaluate our method on a manually annotated data set, and show that our method outperforms the baseline that handles these two tasks separately, boosting the F1 from 80.2% to 83.6% for NER, and the Accuracy from 79.4% to 82.6% for NEN, respectively.
    Page 1, “Abstract”
  2. is trained on a manually annotated data set, which achieves an F1 of 81.48% on the test data set; Chiti-cariu et al.
    Page 3, “Related Work”
  3. We manually annotate a data set to evaluate our method.
    Page 7, “Experiments”
  4. We evaluate our method on a manually annotated data set.
    Page 8, “Conclusions and Future work”

See all papers in Proc. ACL 2012 that mention manually annotated.

See all papers in Proc. ACL that mention manually annotated.

Back to top.

word pair

Appears in 4 sentences as: word pair (3) word pairs (1)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. Hereafter, we use tm to denote the mth tweet ,tfn and to denote the 73th word of of tm and its BIL OU label, respectively, and If; to denote the factor related to 1 and Next, for each word pair with the same lemma, denoted by 753,, and 75%,, we introduce a binary random variable, denoted by 277%”, whose value indicates whether 75%,, and ti, belong to two mentions of the same entity.
    Page 2, “Introduction”
  2. 11We first conduct a simple dictionary—lookup based normalization with the incorrect/correct word pair list provided by Han et al.
    Page 6, “Our Method”
  3. There are two possible ways to fix these errors: 1) Extending the scope of z-serial variables to each word pairs with a common prefix; and 2) developing advanced normalization components to restore such slang expressions and informal abbreviations into their canonical forms.
    Page 8, “Experiments”
  4. One unique characteristic of our model is that a NE normalization variable is introduced to indicate whether a word pair belongs to the mentions of the same entity.
    Page 8, “Conclusions and Future work”

See all papers in Proc. ACL 2012 that mention word pair.

See all papers in Proc. ACL that mention word pair.

Back to top.

graphical model

Appears in 3 sentences as: graphical model (3)
In Joint Inference of Named Entity Recognition and Normalization for Tweets
  1. We propose a novel graphical model to simultaneously conduct N ER and N EN on multiple tweets to address these challenges.
    Page 1, “Abstract”
  2. We propose jointly conducting NER and NEN on multiple tweets using a graphical model , to address these challenges.
    Page 2, “Introduction”
  3. We adopt a factor graph as our graphical model , which is constructed in the following manner.
    Page 2, “Introduction”

See all papers in Proc. ACL 2012 that mention graphical model.

See all papers in Proc. ACL that mention graphical model.

Back to top.

PDF available from www.aclweb.org.
Send us an email