In linguistics, anaphora resolution is the process of determining what noun or real-world object a pronoun is referencing. It may be used to discuss the process by which the human mind understands the reference, or it may mean the statistical models devised by linguists to determine the most likely referent. The term "anaphora resolution" may cause some confusion due to the assortment of possible meanings of the word anaphora, which comes from the Greek "to carry back" or "to carry up." In rhetoric, anaphora is the repetition of key words at the beginning of phrases, sentences, or paragraphs.
Within the discipline of linguistics, anaphora most precisely refers to a third-person pronoun that references back to a noun that has already been mentioned within a text. When discussing anaphora resolution, it may have a more general meaning of any pronoun, whether its referent is found before or after it in the text, or outside the text altogether. For an example of the latter, a person might point to a painting and say, "I like that." In the real-world context, "that" would mean "the painting I'm pointing to."
When only the linguistic content is available for analysis, sometimes there is only one noun in the sentence that matches the anaphora in grammatical person, number, and gender. In this case the anaphora resolution is clear, as in: "Laura is stronger than she looks," with the referent of "she" being "Laura." If there is more than one noun in the sentence, however, the reader relies on common sense to determine the referent, as in: "Some girls wear pink bows in their hair because they like them." In this sentence, "they" clearly references the girls, while "them" clearly references the bows, because bows are not capable of liking girls, but girls are capable of liking bows.
On the other hand, if the sentence were placed in a different context, the reference might not be so clear: "Girls do many things to attract the attention of boys. Some girls wear bows in their hair because they like them." In this case, the "they" may refer to "girls," or it may refer to "boys." An argument might be made for either antecedent.
Linguists have devised statistical algorithms for predicting anaphora resolution using factors such as the distance from pronoun to referent, grammatical agreement of noun with pronoun, and the animaticity of the noun. Some of these algorithms may result in the proper anaphora resolution in the vast majority of cases, but no algorithm in existence has the same level of common sense as an intelligent human reader. The following sentence demonstrates this fact: "The seal pushed the ball with a flipper, then balanced the ball on its nose." Grammatically speaking, "its" could refer to the seal, the flipper, or the ball — but only one of these options actually has a nose. To develop a code that would eliminate all illogical possibilities, such as a ball having a nose, would seem virtually impossible.
