A method and apparatus for summarizing text in a display advertisement are disclosed. A text summary method according to an embodiment of the present invention comprises receiving original data of an advertisement in a text summary method in a display advertisement, and using a pre-learned sequence tag model to determine a property corresponding to a preset attribute from the received original data. tagging the part; calculating an importance for each of the texts of the part tagged by the sequence tag model; and determining a main text of the advertisement for summarizing within a text length preset for the display advertisement based on the calculated importance. Furthermore, the text summarization method according to an embodiment of the present invention is based on language model embedding and word embedding learned in advance using training data and preset product data. The method may further include providing a word embedding characteristic, wherein the tagging may include tagging a portion corresponding to a preset attribute from the received original data by reflecting the language model embedding characteristic and the word embedding characteristic.
展开▼