This post is the simplified version of the papers which I explained previously .(Learning when to split sentences fro natural simplification and Natural simplifiaction for social inclusion)
The papers were very well explained by the author(Sandra Aluisio) when asked for the demonstration for the same with an example and it is quite interesting to see the method they have used for the simplification.
They use SIMPLIFICA(Editor) a tool for producing simple texts in Portuguese and the target audience are teachers,publishers,journalists and many more.
Two kinds of simplification are provided by the tool that is Lexical and Syntactical and a good user interface is provided where we have options to choose between the two .
Before the text is simplified the text is checked for its readability assessment that relies on several features extracted by Coh-Metrix-Port .There are three levels Advanced,Basic and Rudimentary.
when the user clicks the button for readability assessment ,the result is the readability level and the details of that level.
Now we assume our text readabilty is Advanced and we go for simplification by clicking lexical simplification first ,the order of simplification is left to the user
1)Lexical Simplification:
In this process the complex words from the original text is identified and highlighted but the information on how the words are identified as complex is not given in the paper.
They use two dictionaries to find the synonyms i.e TEP and PAPEL ,the user can choose any one of them.
When the user clicks on the word highlighted ,it displays number of synonyms out of which they can choose one.
for ex: 'Claimed' word can be replaced by 'Asked' but in the list we get 'to ask ' as the synonym .For this case ,an edit word option is provided where we can edit the word ,here we edit the word 'to ask ' to 'asked' and then replace.
Sometimes the word can directly be replaced without such difficulties ,for example replacing effort by work.
If there is no synonym found then the word is changed manually.
The above process is repeated for all words and the final revision of the simplified text is obtained which is got by replacing synonyms.
Again the readability level of the text is obtained ,If the text needs more simplification then we go for Syntactical simplification.
2) Syntactic simplification:
Here the simplification is done to each and every sentence.In each sentence few complex phenomena are identified like apposition, relative clauses, coordination, subordination, passive voice, sentences in any order other than subject-verb-object (SVO) and long adverbial phrases.
There are simplification operations to deal with these with these phenomena like split sentence, change ambiguous discourse marker by simpler ones, transform passive to active voice, invert the order of clauses, transform to subject-verb-object ordering, and move long adverbial phrases.
The application of the operation depends on the output of the parser.
There are three kinds of simplification :
1)Personalized: in which he/she can select the syntactic phenomena that he/she wants to have considered for simplification.
2) Strong, in which all complex phenomena considered by the tool will be simplified – this type of simplification generate texts that are adequate for readers at the Rudimentary literacy level
3) Natural, in which a machine-learning system selects which sentences should be simplified so that the resulting text is “natural”, not over simplified, and adequate for readers at the Basic literacy level.
For example here we choose personalized simplification ,
At first the tool marks sentences which needs simplification.
Once the user clicks on it ,the simplified version of the text is obtained .The output has original text,simplified text and corrected simplified version.
Again its user choice to choose the level of simplification.If the user is not happy with the simplification He/she can edit it directly taking the suggested simplifications as advice for how to adapt the sentences (corrected simplified version).
Original:“Crucial to the performance of coach Carlos Alberto Parreira’s “magic square”, Kaká, the biggest highlight of the friendly game against Lucerne’s team, told his attack team-mates:”
Simplified version:“Crucial to the performance of coach Carlos Alberto Parreira’s “magic square” is Kaká. Kaká, the biggest highlight of the friendly game against Lucerne’s team, told his attack team-mates:”
Corrected :“Kaká is crucial to the performance of coach Carlos Alberto Parreira’s “magic square”. Kaká, the biggest highlight of the friendly game against Lucerne’s team, told his attack team-mates:”
Example 2)
Original: “He scored against Fluminense (under-20 team, game-practice match on Sunday) and scored today (yesterday).”
Simplified:“He scored against Fluminense (under-20 team, game-practice match on Sunday). He scored today (yesterday).”
The user can continue the simplification process in order to make the text still more simpler by clicking the Syntactic button.
for ex the first example is chosen for simplification the second time .
Here the second half of the first sentence is chosen for simplification.
Original: “Kaká, the biggest highlight of the friendly game against Lucerne’s team, told his attack team-mates:”
Simplified:“Kaká is the biggest highlight of the friendly game against Lucerne’s team. Kaká told his attack team-mates:”
The final version of the text is obtained and the readability assessment is done again.
As we have seen in this demonstration, the simplification process can help the creation of simple texts that are accessible to people at low literacy levels. Although the output of the simplification process is not always accurate, we consider that the overall tool has a remarkable role in guiding and helping authors to create simple texts.
The classifier a tool for Natural simplification which I mentioned in the previous papers is not integrated in the SIMPLIFICA yet and it is their agenda.
Hope this explanation helps you in better understanding of the concept.
Cya.........
No comments:
Post a Comment