22.1.4 Practical NER Architectures

Commercial approaches to NER are often based on pragmatic combinations of lists, rules and supervised machine learning (Jackson and Moulinier, 2002). One common approach is to make repeated passes over a text allowing the results of one pass to influence the next. The stages typically first involve the use of rules that have extremely high precision but low recall. Subsequent stages employ more error-prone statistical methods that take the output of the first pass into account.

1. First use high-precision rules to tag unambiguous entity mentions;

2. Then search for sub-string matches of the previously detected names using probabilistic string matching metrics (as described in Ch. 19).

3. Consult application-specific name lists to identify likely name entity mentions from the given domain.

4. Finally, apply probabilistic sequence labeling techniques that make use of the tags from previous stages as additional features.

The intuition behind this staged approach is two-fold. First, some of the entity mentions in a text will be more clearly indicative of a given entity's class than others. Second, once an unambiguous entity mention is introduced into a text, it is likely that subsequent shortened versions will refer to the same entity (and thus the same type of entity).

原书第 843 页

Section 22.2. Relation Detection and Classification

| --- | --- | --- | --- |

Figure 22.11 Typical semantic relations with examples and the named entity types they involve.

22.2 RELATION DETECTION AND CLASSIFICATION

Next on our list of tasks is the ability to discern the relationships that exist among the entities detected in a text. To see what this means, let's return to our sample airline text with all the entities marked.

Citing high fuel prices, [ORG United Airlines] said [TIME Friday] it has increased fares by [MONEY $6] per round trip on flights to some cities also served by lower-cost carriers. [ORG American Airlines], a unit of [ORG AMR Corp.], immediately matched the move, spokesman [PERS Tim Wagner] said. [ORG United], a unit of [ORG UAL Corp.], said the increase took effect [TIME Thursday] and applies to most routes where it competes against discount carriers, such as [LOC Chicago] to [LOC Dallas] and [LOC Denver] to [LOC San Francisco].

This text stipulates a set of relations among the named entities mentioned within it. We know, for example, that Tim Wagner is a spokesman for American Airlines, that United is a unit of UAL Corp., and that American is a unit of AMR. These are all binary relations that can be seen as instances of more generic relations such as part-of or employs that occur with fairly high frequency in new-style texts. Fig. 22.11 shows a list of generic relations of the kind used in recent standardized evaluations. $ ^{3} $ More domain-specific relations that might be extracted include the notion of an airline route. For example, from this text we can conclude that United has routes to Chicago, Dallas, Denver and San Francisco.

These relations correspond nicely to the model-theoretic notions we introduced in Ch. 17 to ground the meanings of the logical forms. That is, a relation

原书第 844 页

| Domain\nUnited, UAL, American Airlines, AMR\nTim Wagner\nChicago, Dallas, Denver, and San Francisco | $ \mathcal{D} = \{a, b, c, d, e, f, g, h, i\} $\na,b,c,d\ne\nf,g,h,i |

| --- | --- |

| Classes\nUnited, UAL, American and AMR are organizations\nTim Wagner is a person\nChicago, Dallas, Denver and San Francisco are places | $ Org = \{a, b, c, d\} $\n $ Pers = \{e\} $\n $ Loc = \{f, g, h, i\} $ |

| Relations\nUnited is a unit of UAL\nAmerican is a unit of AMR\nTim Wagner works for American Airlines\nUnited serves Chicago, Dallas, Denver and San Francisco | $ PartOf = \{\langle a, b \rangle, \langle c, d \rangle\} $\n $ OrgAff = \{\langle c, e \rangle\} $\n $ Serves = \{\langle a, f \rangle, \langle a, g \rangle, \langle a, h \rangle, \langle a, i \rangle\} $ |

Figure 22.12 A model-based view of the relations and entities in our sample text.

consists of set of ordered tuples over elements of a domain. In most standard information extraction applications, the domain elements correspond either to the named entities that occur in the text, to the underlying entities that result from co-reference resolution, or to entities selected from a domain ontology. Fig. 22.12 shows a model-based view of the set of entities and relations that can be extracted from our running example. Notice how this model-theoretic view subsumes the NER task as well; named entity recognition corresponds to the identification of a class of unary relations.