Sentence summarization aims to shorten a given sentence and produce a 
brief summary of it. The first part of my presentation is about modeling 
syntactic feature for sentence summarization. Features can be word-level 
or structural sentence-level. These features can be added to source 
texts or output summaries. I show that modeling parse information of the 
input sentences by linearization of PCFG trees generates better 
summaries, more similar to human-written ones. It solve the repetition 
problem-generating a nonsense sequence of a word or phrase. Our manual 
evaluation based on readability and informativeness shows that this 
model outperforms the baseline in both criteria.
In the next part of the presentation, I briefly talk about an ongoing 
study which I call "dynamic processing of text". I explain a model which 
is able to dynamically change the encoding process of the input text. 
The model can filter out unnecessary information to produce a more 
accurate representation of the text.