Tutorial4-Further_analysis_and_configuration_of_Concept_Network

Introduction

This tutorial goes into further details demonstrating the use of the package, covering functions in r/02_data_exploration.R and r/03_concept_network.R. The intention is to provide an example where some initial analysis of an open-ended survey question is used to inform the settings and concept words of the Concept Network plot. Then, we demonstrate how the Concept Network settings can be fine-tuned to produce plots which are more persuasive visualisations of your findings.

Installation of package.

Once the package is installed, you can load the finnsurveytext package as below: (Other required packages such as dplyr and stringr will also be installed if they are not currently installed in your environment.)

library(finnsurveytext)

Overview of Functions

The functions covered in this tutorial are:

The Data

We’re looking again at the FSD2821 Young People’s Views on Development Cooperation 2012 (FSD2821 Nuorten ajatuksia kehitysyhteistyöstä 2012) survey which is provided as sample data with the package.

Development Cooperation Data

data/conllu_dev_q11_1.rda
data/conllu_dev_q11_1_nltk.rda
data/conllu_dev_q11_1_snow.rda
data/conllu_dev_q11_2.rda
data/conllu_dev_q11_2_nltk.rda
data/conllu_dev_q11_3.rda
data/conllu_dev_q11_3_nltk.rda

There are 3 open-ended questions we look at within this data and we will set the threshold and other function arguments to be suit each question.

The open-ended questions

q11_1 Jatka lausetta: Kehitysmaa on maa, jossa… (Avokysymys)
- q11_1 Continue the sentence: A developing country is a country where… (Open question)
q11_2 Jatka lausetta: Kehitysyhteistyö on toimintaa, jossa… (Avokysymys)
- q11_2 Continue the sentence: Development cooperation is an activity in which… (Open question)
q11_3 Jatka lausetta: Maailman kolme suurinta ongelmaa ovat… (Avokysymys)
- q11_3 Continue the sentence: The world’s three biggest problems are… (Open question)

Demonstration of analysis

Read in

First, we read in our data.

devdata <- dev_data

q11_1 <- conllu_dev_q11_1_nltk
q11_2 <- conllu_dev_q11_2_nltk
q11_3 <- conllu_dev_q11_3_nltk

Here we’ll print the first 10 rows of the the raw data and then one of the the processed datasets (Q11_1).

knitr::kable(head(dev_data, 5))

fsd_id	q11_1	q11_2	q11_3
1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka). Ja niitä ei ole riittävästi	varmaan koitetaan kehittää em	saastuminen ja luonnonvarojen liikakäyttö, nälänhätä ja ylikansoittuminen
2	on kurjuutta ja nälänhätää, asiat eivät ole vielä kehittyneet, lapset eivät pääse kouluun, tytöillä on huonompi asema kuin pojilla.	pyritään auttamaan?	ihmiskauppa, nälänhätä ja sodat/turvattomuus
3	jokaisella ei ole turvattua toimeentuloa ja jossa todella huomaa koulutuksen arvon.	autetaan ja näytetään ihmisille tie parempaan tulevaisuuteen heidän oman työnsä tuloksena.	kouluttamattomuus, nälkä ja puhtaan veden puute.
4	kehityksen taso ei ole yhtä korkea kuin kehittyneissä maissa	yleensä haitataan kehitysmaan kehittymistä	öljy, raha, se fakta että ei olla vielä päästy asumaan muualla kuin tällä yhdellä planeetalla
5	asiat ovat vielä huonossa jamassa ja apua tarvitaan		eriarvoisuus, sodat, nälänhätä tietyissä maissa


knitr::kable(head(q11_1, 10))

doc_id	paragraph_id	sentence_id	sentence	token_id	token	lemma	upos	xpos	feats	head_token_id	dep_rel	deps	misc
doc1	1	1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka).	1	elämiseen	eläminen	NOUN	N,Sg,Ill	Case=Ill\|Number=Sing	2	nmod	NA	NA
doc1	1	1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka).	2	tarvittavat	tarvita	VERB	V,Pass,PcpVa,Pl,Nom	Case=Nom\|Number=Plur\|PartForm=Pres\|VerbForm=Part\|Voice=Pass	3	acl	NA	NA
doc1	1	1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka).	3	perusasiat	perusasia	NOUN	N,Pl,Nom	Case=Nom\|Number=Plur	5	nsubj	NA	NA
doc1	1	1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka).	5	kehittymättöneet	kehittymättöa	VERB	V,Act,PcpNut,Pl,Nom	Case=Nom\|Number=Plur\|PartForm=Past\|VerbForm=Part\|Voice=Act	0	root	NA	NA
doc1	1	1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka).	7	esim.	esim.	PART	Pcle,Abbr	Abbr=Yes	8	dep	NA	NA
doc1	1	1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka).	8	vesi	vesi	NOUN	N,Sg,Nom	Case=Nom\|Number=Sing	5	obj	NA	SpaceAfter=No
doc1	1	1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka).	10	talo	talo	NOUN	N,Sg,Nom	Case=Nom\|Number=Sing	8	conj	NA	SpaceAfter=No
doc1	1	1	elämiseen tarvittavat perusasiat ovat kehittymättöneet (esim. vesi, talo, ruoka).	12	ruoka	ruoka	NOUN	N,Sg,Nom	Case=Nom\|Number=Sing	8	conj	NA	SpaceAfter=No
doc1	1	2	Ja niitä ei ole riittävästi	5	riittävästi	riittävästi	ADV	Adv	NA	4	advmod	NA	SpacesAfter=
doc2	1	1	on kurjuutta ja nälänhätää, asiat eivät ole vielä kehittyneet, lapset eivät pääse kouluun, tytöillä on huonompi asema kuin pojilla.	2	kurjuutta	kurjuut	NOUN	N,Sg,Par	Case=Par\|Number=Sing	0	root	NA	NA

Initial EDA

To begin with, let’s look at the differences between the types of reponses we get for each of these open-ended questions by using some of the EDA functions. (For further details into these functions, see “Tutorial2-Data_Exploration”)

First, let’s consider these functions:

fst_summarise() - This table will indicate response rate, and word counts.
fst_pos() - This table counts the number and proportion of words with each part-of-speech tag.
fst_length_summary() - This table gives the average and quartile values of lengths of responses in words and sentences.

The differences in the types of responses can be seen in the exploratory data analysis results below:

Summarise

knitr::kable(
  fst_summarise(data = q11_1, desc = "Q11_1")
)

Description	Respondents	No Response	Proportion	Total Words	Unique Words	Unique Lemmas
Q11_1	945	24	0.98	4257	1513	1065

knitr::kable(
  fst_summarise(data = q11_2, desc = "Q11_2")
)

Description	Respondents	No Response	Proportion	Total Words	Unique Words	Unique Lemmas
Q11_2	941	34	0.97	4407	1270	893

knitr::kable(
  fst_summarise(data = q11_3, desc = "Q11_3")
)

Description	Respondents	No Response	Proportion	Total Words	Unique Words	Unique Lemmas
Q11_3	945	25	0.97	4192	1132	994

Remarks:

All 3 questions have a high (97/98%) response rate.
Q11_1 contains the most unique words.
All 3 questions have a similar number of unique lemmas.

POS Summary

knitr::kable(
  fst_pos(data = q11_1)
)

UPOS	UPOS_Name	Count	Proportion
ADJ	adjective	664	0.156
ADP	adposition	65	0.015
ADV	adverb	399	0.094
AUX	auxiliary	29	0.007
CCONJ	coordinating conjunction	4	0.001
DET	determiner	79	0.019
NOUN	noun	2225	0.523
NUM	numeral	2	0.000
PART	particle	160	0.038
PRON	pronoun	47	0.011
PROPN	proper noun	18	0.004
SCONJ	subordinating conjunction	4	0.001
SYM	symbol	2	0.000
VERB	verb	552	0.130
X	other	7	0.002

knitr::kable(
  fst_pos(data = q11_2)
)

UPOS	UPOS_Name	Count	Proportion
ADJ	adjective	353	0.080
ADP	adposition	55	0.012
ADV	adverb	178	0.040
AUX	auxiliary	20	0.005
CCONJ	coordinating conjunction	4	0.001
DET	determiner	66	0.015
INTJ	interjection	2	0.000
NOUN	noun	1974	0.448
NUM	numeral	3	0.001
PART	particle	62	0.014
PRON	pronoun	20	0.005
PROPN	proper noun	7	0.002
SCONJ	subordinating conjunction	11	0.002
SYM	symbol	1	0.000
VERB	verb	1648	0.374
X	other	3	0.001

knitr::kable(
  fst_pos(data = q11_3)
)

UPOS	UPOS_Name	Count	Proportion
ADJ	adjective	389	0.093
ADP	adposition	24	0.006
ADV	adverb	64	0.015
AUX	auxiliary	3	0.001
CCONJ	coordinating conjunction	3	0.001
DET	determiner	28	0.007
INTJ	interjection	2	0.000
NOUN	noun	3311	0.790
NUM	numeral	5	0.001
PART	particle	29	0.007
PRON	pronoun	12	0.003
PROPN	proper noun	31	0.007
SYM	symbol	1	0.000
VERB	verb	278	0.066
X	other	12	0.003

Remarks:

Nouns are the most common word type across all 3 questions.
Q11_1 also commonly contains adjectives, adverbs and nouns.
Q11_2 has a similar number of nouns and verbs, and then a smaller number of adjectives and adverbs.
Q11_3 also has a lot of adjectives and verbs but few adverbs.

Length Summary

knitr::kable(
  fst_length_summary(
    data = q11_1,
    desc = "Q11_1",
    incl_sentences = TRUE
  )
)

Description	Respondents	Mean	Minimum	Q1	Median	Q3	Maximum
Q11_1- Words	921	6.514658	1	3	5	8	33
Q11_1- Sentences	921	1.057546	1	1	1	1	4

knitr::kable(
  fst_length_summary(
    data = q11_2,
    desc = "Q11_2",
    incl_sentences = TRUE
  )
)

Description	Respondents	Mean	Minimum	Q1	Median	Q3	Maximum
Q11_2- Words	907	5.398015	1	3	4	7	30
Q11_2- Sentences	907	1.018743	1	1	1	1	3

knitr::kable(
  fst_length_summary(
    data = q11_3,
    desc = "Q11_2",
    incl_sentences = TRUE
  )
)

Description	Respondents	Mean	Minimum	Q1	Median	Q3	Maximum
Q11_2- Words	920	5.515217	1	4	5	6	32
Q11_2- Sentences	920	1.015217	1	1	1	1	3

Remarks:

All 3 questions have most responses being one sentence.
Each question has a longest response of about 30 words and most responses of about 5 words.

Most common words and n-grams

Now, lets create some tables of the most common words, bigrams and trigrams using the following functions:

fst_freq()
fst_ngrams()

Top Words

fst_freq(
  data = q11_1,
  number = 10,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_1"
)
#> Note:
#>  Words with equal occurrence are presented in alphabetical order. 
#>  By default, words are presented in order to the `number` cutoff word. 
#>  This means that equally-occurring later-alphabetically words beyond the cutoff word will not be displayed.


fst_freq(
  data = q11_2,
  number = 10,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_2"
)
#> Note:
#>  Words with equal occurrence are presented in alphabetical order. 
#>  By default, words are presented in order to the `number` cutoff word. 
#>  This means that equally-occurring later-alphabetically words beyond the cutoff word will not be displayed.


fst_freq(
  data = q11_3,
  number = 10,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_3"
)
#> Note:
#>  Words with equal occurrence are presented in alphabetical order. 
#>  By default, words are presented in order to the `number` cutoff word. 
#>  This means that equally-occurring later-alphabetically words beyond the cutoff word will not be displayed.

(We leave the default for norm her so “occurrence” is based on the number of words in the set of responses for each question.)

Here we can see that there is a clear ‘top’ set of words in each question:

Q11_1: ‘ihminen’ (man)
Q11_2: ‘kehitysmaa’ (developing country), ‘auttaa’ (helps)
Q11_3: ‘köyhyys’ (poverty), ‘nälänhätä’ (famine), ‘sota’ (war)

Bi-grams and Trigrams

Here we look at common sets of two or three words.

fst_ngrams(
  data = q11_1,
  number = 10,
  ngrams = 2,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_1"
)
#> Note:
#>  N-grams with equal occurrence are presented in alphabetical order. 
#>  By default, n-grams are presented in order to the `number` cutoff n-gram. 
#>  This means that equally-occurring later-alphabetically n-grams beyond the cutoff n-gram will not be displayed.

fst_ngrams(
  data = q11_1,
  number = 10,
  ngrams = 3,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_1"
)
#> Note:
#>  N-grams with equal occurrence are presented in alphabetical order. 
#>  By default, n-grams are presented in order to the `number` cutoff n-gram. 
#>  This means that equally-occurring later-alphabetically n-grams beyond the cutoff n-gram will not be displayed.

fst_ngrams(
  data = q11_2,
  number = 10,
  ngrams = 2,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_2"
)
#> Note:
#>  N-grams with equal occurrence are presented in alphabetical order. 
#>  By default, n-grams are presented in order to the `number` cutoff n-gram. 
#>  This means that equally-occurring later-alphabetically n-grams beyond the cutoff n-gram will not be displayed.

fst_ngrams(
  data = q11_2,
  number = 10,
  ngrams = 3,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_2"
)
#> Note:
#>  N-grams with equal occurrence are presented in alphabetical order. 
#>  By default, n-grams are presented in order to the `number` cutoff n-gram. 
#>  This means that equally-occurring later-alphabetically n-grams beyond the cutoff n-gram will not be displayed.

fst_ngrams(
  data = q11_3,
  number = 10,
  ngrams = 2,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_3"
)
#> Note:
#>  N-grams with equal occurrence are presented in alphabetical order. 
#>  By default, n-grams are presented in order to the `number` cutoff n-gram. 
#>  This means that equally-occurring later-alphabetically n-grams beyond the cutoff n-gram will not be displayed.

fst_ngrams(
  data = q11_3,
  number = 10,
  ngrams = 3,
  norm = "number_words",
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  strict = TRUE,
  name = "Q11_3"
)
#> Note:
#>  N-grams with equal occurrence are presented in alphabetical order. 
#>  By default, n-grams are presented in order to the `number` cutoff n-gram. 
#>  This means that equally-occurring later-alphabetically n-grams beyond the cutoff n-gram will not be displayed.

Recall the open-ended questions

q11_1 Jatka lausetta: Kehitysmaa on maa, jossa… (Avokysymys)
- q11_1 Continue the sentence: A developing country is a country where… (Open question)
q11_2 Jatka lausetta: Kehitysyhteistyö on toimintaa, jossa… (Avokysymys)
- q11_2 Continue the sentence: Development cooperation is an activity in which… (Open question)
q11_3 Jatka lausetta: Maailman kolme suurinta ongelmaa ovat… (Avokysymys)
- q11_3 Continue the sentence: The world’s three biggest problems are… (Open question)

Common themes raised in bigrams and trigrams:

Q11_1
- Discussion about groups of people living in poverty
- ‘suuri osa ihminen’, ‘tarvita apu’, ‘ihminen nähdaä nälkä’, ‘osa ihminen elää’, ihminen elää köyhyysraja’, ‘köyhä ihminen’
Q11_2
- Themes about helping countries develop
- ‘auttaa kehitysmaa’, ‘pyrkiä auttaa kehitysmaa’, ‘auttaa kehitysmaa auttaa’
Q11_3
- Concern about lack of water, clean water
- ‘vesi puute’, ‘puhdas vesi’, ‘puhdas vesi puute’

Wordcloud

fst_wordcloud()

We can see the most frequent words identified above also coming out in the wordclouds.

fst_wordcloud(
  data = q11_1,
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  max = 100
)


fst_wordcloud(
  data = q11_2,
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  max = 100
)


fst_wordcloud(
  data = q11_3,
  pos_filter = c("NOUN", "VERB", "ADJ", "ADV"),
  max = 100
)

Exploring some concept networks based on thematic frequently occurring words

fst_concept_network()

Based on the most frequently-occurring words, we have chosen some lists of “concepts” for each question and will create Concept Networks based on these. You can see below how the threshold can be used to make sure the concept network isn’t too “busy” by removing less frequent connection words. Similarly, you can see how the length of the “concepts” list impacts the “busyness” of the plot. Often, the “concept” list and appropriate threshold will be the product of trial and error.

Q11_1

q11_1 Jatka lausetta: Kehitysmaa on maa, jossa… (Avokysymys)
- q11_1 Continue the sentence: A developing country is a country where… (Open question)
Chosen concept words : “elintaso, köyhä, ihminen”
- (In English) standard of living, poor, man

fst_concept_network(
  data = q11_1,
  concepts = "elintaso, köyhä",
  threshold = NULL,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_1 - No threshold"
)

fst_concept_network(
  data = q11_1,
  concepts = "elintaso, köyhä",
  threshold = 3,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_1 - Threshold = 3"
)

fst_concept_network(
  data = q11_1,
  concepts = "elintaso, köyhä, ihminen",
  threshold = NULL,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_1 - No threshold"
)

fst_concept_network(
  data = q11_1,
  concepts = "elintaso, köyhä, ihminen",
  threshold = 3,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_1 - Threshold = 3"
)

Remarks:

Here, our first plot may be the best. We can see that a threshold is not required as there are not too many words displayed but we can get some insight into the use of these words (in English, they are ‘standard-of-living’ and ‘poor’). As you can see in plots 3 and 4, when including the most common word, ‘ihminen’ (‘human being’ or ‘man’), a threshold such as 3 is advisable.

Q11_2

q11_2 Jatka lausetta: Kehitysyhteistyö on toimintaa, jossa… (Avokysymys)
- q11_2 Continue the sentence: Development cooperation is an activity in which… (Open question)
Chosen concept words: “kehitysmaa, auttaa, pyrkiä, maa, ihminen”
- (In English) development, help, strive, country, man

fst_concept_network(
  data = q11_2,
  concepts = "kehitysmaa, auttaa, pyrkiä, maa, ihminen",
  threshold = NULL,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_2 - No threshold"
)

fst_concept_network(
  data = q11_2,
  concepts = "kehitysmaa, auttaa, pyrkiä, maa, ihminen",
  threshold = 10,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_2 - Threshold = 10"
)

fst_concept_network(
  data = q11_2,
  concepts = "kehitysmaa, auttaa, pyrkiä, maa, ihminen",
  threshold = 5,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_2 - Threshold = 5"
)


fst_concept_network(
  data = q11_2,
  concepts = "kehitysmaa, auttaa, pyrkiä",
  threshold = NULL,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_2 - No threshold"
)

fst_concept_network(
  data = q11_2,
  concepts = "kehitysmaa, auttaa, pyrkiä",
  threshold = 5,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_2 - Threshold = 5"
)

Remarks:

In Q11_2, plots 1-3 show that if we include all 5 of our words (‘developing country’, ‘help’, ‘strive’, ‘country’, ‘man’) or a subset that a threshold is required, but that a threshold of 10 is too large to gain additional words in the Network. Here threshold = 5 seems appropriate.

Q11_3

q11_3 Jatka lausetta: Maailman kolme suurinta ongelmaa ovat… (Avokysymys)
- q11_3 Continue the sentence: The world’s three biggest problems are… (Open question)
Chosen conpcet words: “köyhyys, nälänhätä, sota, ilmastonmuutos, puute”
- (In English) poverty, famine, war, climate change, lack of

fst_concept_network(
  data = q11_3,
  concepts = "köyhyys",
  threshold = NULL,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_3 - köyhyys / No threshold"
)

fst_concept_network(
  data = q11_3,
  concepts = "köyhyys, puute",
  threshold = NULL,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_3 - köyhyys, puute / No threshold"
)

fst_concept_network(
  data = q11_3,
  concepts = "köyhyys, nälänhätä, sota, ilmastonmuutos, puute",
  threshold = NULL,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_3 - No threshold"
)

fst_concept_network(
  data = q11_3,
  concepts = "köyhyys, nälänhätä, sota, ilmastonmuutos, puute",
  threshold = 3,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_3 - Threshold = 3"
)

fst_concept_network(
  data = q11_3,
  concepts = "köyhyys, nälänhätä, sota, ilmastonmuutos, puute",
  threshold = 2,
  norm = "number_words",
  pos_filter = NULL,
  title = "Q11_3 - Threshold = 2"
)

Remarks:

In Q11_3, we can see that if your concept word list is short (such as a single word) thresholds generally are not required, but that with longer concept lists, setting a threshold = 2 or threshold = 3, we can simplify and improve a plot that is a little too crowded in this case. The “best” threshold is generally a matter of context and at the analyst’s discretion.

Conclusion

From the above, we can see that different settings create different Concept Networks. There is no “right” setting for a Concept Network, so it is worthwhile exploring the Concept Networks that result from different concept words and thresholds to investigate the data and identify trends which warrant further analysis. Initial Concept Network settings can be informed by other functions in finnsurveytext, such as choosing the most frequent words/n-grams or considering insights from wordclouds.

Citation

Finnish Children and Youth Foundation: Young People’s Views on Development Cooperation 2012 [dataset]. Version 2.0 (2019-01-22). Finnish Social Science Data Archive distributor]. http://urn.fi/urn:nbn:fi:fsd:T-FSD2821

Tutorial4-Further_analysis_and_configuration_of_Concept_Network_functions

Introduction

Installation of package.

Overview of Functions

The Data

Development Cooperation Data

The open-ended questions

Demonstration of analysis

Read in

Initial EDA

Summarise

POS Summary

Length Summary

Most common words and n-grams

Top Words

Bi-grams and Trigrams

Wordcloud

Exploring some concept networks based on thematic frequently occurring words

Q11_1

Q11_2

Q11_3

Conclusion

Citation