Page 178 - The-5th-MCAIT2021-eProceeding

4. Results and Discussions

4.1 Computing Summary Output Area
Using the Opencv image processing library, the length and width of the summary output slot is obtain. The
width and length of this portion is used to compute the area of the shape of the portion which will be used to
determine the maximum length of the summary. In Figure 4.1, the maroon portion within the image is the
designated slot where the summary will be displayed. Based on the image above the area of the portion is 0.36in
which will be used as the maximum length of the output summary therefore making sure that summary fits the
portion perfectly.

Fig. 3. Detected and analysed summary output area.

4.2. Generating Summary
The summary is generated by modifying the T5 (Text-To-Text Transfer Transformer) model and adding the
area constraint. T5 is a new transformer model from Google that is trained in an end-to-end manner with text
as input and modified text as output (Raffel et al., 2020). First, an input sequence of tokens is mapped to a
sequence of embeddings, which is then passed into the encoder. The encoder consists of a stack of “blocks”,
each of which comprises two subcomponents: a self-attention layer followed by a small feed-forward network.
Layer normalization (Ba et al., 2016) is applied to the input of each subcomponent. A simplified version of
layer normalization is used where the activations are only rescaled and no additive bias is applied. After layer
normalization, a residual skip connection adds each subcomponent’s input to its output.

4.3. Length-controllable Summary

The computed area is used to obtain the minimum and maximum length of the summary and; these will be
parsed in T5 model to generate an abstractive summary that fits the summary output slot perfectly. Using Figure
4.1, the area of 0.36in is parsed to the T5 model which generate then generates a summary that fits the specified
output portion. The summary generated without output area constraint does not fit the designated slot of the
summary. Meanwhile, the summary generated with the proposed area constraint model generates summary that
fits desired portion.

E- Proceedings of The 5th International Multi-Conference on Artificial Intelligence Technology (MCAIT 2021) [165]
Artificial Intelligence in the 4th Industrial Revolution