H.M. the Queen's New Year's Address Analyzer

● Options to set prefference to update featured words from topic model.

■ You can more easily inspect topic in other tabs by quicly obtaining their values.

Notice: "Update on topic selection" will clear previous selections.

● Number of terms to display sets the number of terms used in the topic model, by size of topic.

■ Limiting the number of terms makes the topic model easier to interpret. Adding more, gives more information.

● Topic model is a tool that maps topics. It does so by analyzing word relations by analyzing all speeches and mapping the relations as word matrixes. Then the model displays the topics to the left, and the words within on the right. Frequency of use is depicted by size for topics, and bar length for words. Topics are named 1:n, where 1 is the largest topic.

■ The topics model can be used to get an understanding of themes present in the speeches throughout the years.

● A slider to set the maxium amount of sample sentences.

■ To adjust how many sentece samples you want to see.

● Samples of sentences belonging to chosen topic.

■ To see examples of sentences in chosen topic.

● This boxplot shows the average sentiment of sentences belonging to a specific topic, compared to the rest of the topics.

■ To compare a specific topics sentiment to the rest of the topics.

● Sentiment of words displays words by sentiment.

■ Sentiment of words can help you identify what words influenced the sentiment.

Notice: Words with a polarity of 0 is filtered.

● Frequency used displays words by frequency.

■ Frequency used can help you identify what words influenced the sentiment.

Notice: Words with a polarity of 0 is filtered.

● Number of words is a slider to set the number of featured words.

■ Limiting the amount words featured will reduce clutter and make Sentiment of words and Frequency used easier to interpret.

● A world map, with countries mentioned in speeches.

■ To know where countries are, and how often they are mentioned.

● A bar chart showing the amount of times a country is mentioned for each year.

● A boxplot showing the average sentiment of sentences in which a country is mentioned

■ To compare the mentions of a specific country, to mentions of countries in general

● Samples of mentions.

■ To see examples of mentions.

● A slider to set the maxium amount of sample sentences.

■ To adjust how many sentece samples you want to see.

● Top frequent numbers is a slider to filter for frequency of words in word themed visualizations.

■ Many words can be mentioned. Most of these visualizations can become to cluttered to interpret.

Stopwords are filtered. This is done to avoid the most common words (like "the") to dominate the statistics. This is done with stopwords defined by Bertel Torp, stopwords defined by snowball and custom stopwords found from interacting with data.

Words have been stemmed, to get better data for topics. This does remove information about word forms. this is done with snowballC for R

Words have been lemmatized (replacing words with identical meaning with a headword), to improve topic analysis, by using udpipe for R.

Topics have been derived with a Structural Topic Model (STM) that is characterized by estimating topic distributions using a covariate matrix as well as a document-term matrix.

The covariates used for estimating topic distributions in the Queen's new year's addresses is year of speech and average sentiment on sentence level.

For danish: Sentiment analysis is done by getting sentiment values from "Det Danske Sprog- og Litteraturselskab (DSL, Society for Danish Language and Literature) and Center for Sprogteknologi, Københavns Universitet (CST, Centre for Language Technology, University of Copenhagen)", and adding the the values to the words we collected on word level, stem level and lemmatized values.

For danish: Sentiment analysis is done by getting sentiment values from "Det Danske Sprog- og Litteraturselskab (DSL, Society for Danish Language and Literature) and Center for Sprogteknologi, Københavns Universitet (CST, Centre for Language Technology, University of Copenhagen)", and adding the the values to the words we collected on word level, stem level and lemmatized values.

For english: Sentiment analysis is done by getting sentiment values from the AFINN-111 sentiment dataset and adding to the words we collected on word-, stem- and lemmatized-level.

● This are the speeches that are covered in the dashboard, with your current filter.

■ You can use this to see the source from what you filtered.

Notice: English version of speech are translations. Not all Danish speeches are translated. Some years are therefore not available.

If you wish to have some guidance in your data journey, we have prepared assisting labels for each visualization.

These labels will be located directly beneath the point of interest.

---

Click the icon in the top right cornor to hide/expand section.

● What marking: A blue dot with bold dark-blue text indicates what kind of visualization/information is depicted.

■ Why marking: A Yellow circle with dark-yellow bold text indicates why and when to use you could use the visualization/information you can observe.

Grey italic/slanted text indicates other kinds of help text. This could be descriptions on how operate inputs, information about limitations and bugs.

Black text is other kinds of information. It would usually be practical information.

● This dashboard utilizes filtering systems. The main filters are in the sidebar when expanded. Some visualizations have custom filters. They are documented when needed.

■ The filters can be used to filter the data you see. You can do this to inspect a certain subset of data. E.g., to look data relevant to you.

▲ To expand the sidebar, press the expand button in the top left corner.

● The language filter set's what language/translation of speeches you inspect, and what language to display the Wikipedia info box.

■ For this dashboard, Danish would be optimal. It is the original language of the speeches. If you don't understand Danish, you can still use a limited set of speeches in English.

▲ Push the text or dot next to the language you want to use.

● Year filter refers to the group of filters affecting year. These are: the year input, the year slider and the year selector.

■ Filtering years can help you limit the data relevant to the topic you are researching.

▲ Year input: Push the text or dot next to the year input method you need or prefer

▲ Year range: select the start year and end year by drag 'n drop the round dots.

▲ Year selector - remove year(s): click on a year and press Backspace-key or Delete-key to remove a year. Hold down Ctrl to select multiple. Alternatively, select input-field (e.g. by clicking or using tab button), and navigate the cursor with the arrow keys, and use Backspace-key or Delete-key to remove years.

▲ Year selector - Add year: select input-field (e.g. by clicking or using tab button) and type year. Press enter to add or select suggestion by clicking on it.

Notice: Some visualizations will use the whole dataset as a reference regardless of your set filter.

● Featured words is a selection of words from words featured in the speeches, that is used to filter or feature words in different kinds of ways.

■ When you are inspecting a topic, some words might seem important. So, you can filter for these words using the filter.

▲ Select word: Type word and press enter when done or click suggestion pop up.

▲ Remove word(s): click on a word and press Backspace-key or Delete-key to remove word Hold down Ctrl to select multiple. Alternatively, select input-field (e.g. by clicking or using tab button), and navigate the cursor with the arrow keys, and use Backspace-key or Delete-key to remove words.

▲ Remove words by filter: Click button labled: "Clear words"

▲ Regret removal by clear words: Click button labled: "Regret clear". It only works for last clearning.