-
Notifications
You must be signed in to change notification settings - Fork 0
/
guide.txt
68 lines (57 loc) · 3.14 KB
/
guide.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
Guide to the sentiment analyser:
##################################################################################
Evaluated sentiment:
##########################################
Click the buttons to select which sentiment to analyse for.
The selection is preserved through predictions so it does not need to be
re-selected for each prediction
##########################################
Grouping of results:
##########################################
Unsorted:
The evaluated sentences will be displayed togeather with their corresponding
prediction values in chronological order. Input format will not be preserved.
Sorted:
The evaluated sentences will be displayed in the order or prediction magnitude,
from largest to smallest. Input format will not be preserved.
Max:
Each input string is displayed togeather with the maximum prediction of the each
string. E.g if a paragraph is placed in the first cell of a CSV-file and used as
the input. The entire paragraph will be returned with the single highest prediction
that is produced from it. If the free text input is used the entire input will be
treated as a single string. The input format is preserved.
The selection is preserved through predictions so it does not need to be
re-selected for each prediction
##########################################
Input:
##########################################
The model will evaluate all inputs sentece wise. This means that your input text
will be split into sentences based on the presence of "." tokens. This means that
you can use entire paragaphs or documents as input but it will automatically be
split into sentences.
The Bert modes has an maximum input length of 512 input tokens. This translates
to a maximum allowed sentence length of roughly 500 words. Sentences longer than
~500 words cannot be evaluated.
If a file is used as the input it need to be formatted in a CSV-format where the
data to be analyzed is located in the first column. It is expected that the
column has no header. Press select files and chose a CSV-file to evaluate,
only one file can be evaluated at a time.The maximum allowed file size is
10MB but the site is optimized for smaller file sizes, "larger" files will lead
to long prediction times.
The free text input will allways take priority over a file input.
##########################################
Predict:
##########################################
When you have selected what sentiment to evaluate for, how you want the results
presented, and entered an input or uploaded a file. You are ready to press the
"Predict" button and start your evaluation. You will be prompted with se prompt
saying "Processing the data", when this prompt dissapears tour evaluation is complete.
##########################################
Download the results:
##########################################
You can choose to have your results downloaded as either a JSON or CSV file.
The foratting of these files will be based on the selection you made in
"Grouping of results".
For the CSV-file option all commas(",") in output will the changed to
semicolon(";"). This is done to not break the inheret formatting of the
CSV-file (Comma Seperated Variable).