Page 201 - Journal of Library Science in China, Vol.47, 2021

200 Journal of Library Science in China, Vol.13, 2021

Figure 8. Humanities sentiment terms retrieval based on “ search term by the poem ” mode (example)

hand, also refines the poet’s sympathy and lament for this tragedy.
In Figure 9, the terms retrieval mode of “search poem by the term” is further explored based on the
overall knowledge of ancient poetry sentiment. For example, by using “颠沛流离” as the search term,
we can find fifteen poems such as 《野老》《别薛华》《与东吴生相遇》《茅屋为秋风所破
歌》, etc. Most of these poems reflect the poet’s thoughts and feelings of suffering from displacement
and longing for shelter and peace. By using “触景伤情” as the search term, ten ancient poems were
found, including 《夕阳楼》《哀江头》《与诸子登岘山》 and 《秋兴八首选四》, most of
which blend sentiment with scenery and take various types of imagery as a carrier to convey the poet’s
sadness. By extension, using the dual terms “颠沛流离” and “触景伤情” to narrow the retrieval
scope, poems that combine the two types of sentiment poems, like 《兼示符离及下邽弟妹》and
《南征》 are obtained. In the former, the poet and his separated brothers are depicted as lonely and
displaced as the lonely goose that strays from the flock, triggering nostalgia for their loved ones. The
latter depicts the poet sighing about his lack of prospects and bleak prospects in the face of the beauty
in front of him and expresses the sadness for long years of displacement and travel in the South.

4.2 Humanity sentiment granularity mining

Humanity sentiment granularity mining refers to parsing the sentiment connotation of multiple
granularities from different levels of texts. The sentiment knowledge granularity of ancient poems
is expressed as “poem title” > “poem text” > “appreciation” and refined layer by layer, poem “述怀”
will be taken as an example to extract the sentiment terms from the three types of texts. After that,