3. Clustering & Topic Modeling to Discover Higher-Order Patterns of Meaning - [E3] Barron, Alexander TJ, Jenny Huang, Rebecca L. Spang, and Simon DeDeo.

[lkcao]

Post questions here for this week's exemplary readings:

3. Barron, Alexander TJ, Jenny Huang, Rebecca L. Spang, and Simon DeDeo. 2018. “Individuals, institutions, and innovation in the debates of the French Revolution.” Proceedings of the National Academy of Sciences 115(18): 4607-4612.

[cty20010831]

I personally think this is a great paper that applies topic modeling and measures based on information theory to quantify the word-use patterns. While I am impressed by the multi-level analysis of the relationship between novelty, transience, and resonance (individual and group levels), I am wondering if there is the possibility that some people's word-use patterns are influenced by leading figures mentioned in the paper and, if so, how can we quantify and tease them apart?

[ana-yurt]

This paper diachronically employs LDA to measure three key constructs: novelty, transience, and resonance. Besides a system-wide novelty bias, the paper identifies different institutional roles in pushing for or suppressing novelty. Their quantitative findings align with the conventional consensus that the left wing, the bourgeoisie, and speakers on behalf of a committee display high novelty, while right-wing speakers are low-novelty. Moreover, the abnormally high-resonance speakers, such as Robespierre and Pétion, match the conventional wisdom of key players in this history.

An interesting conclusion is that on the power of committees. Relying on prior literature and qualitative reading, the authors somewhat conveniently explained the strange pattern of a decreasing new-item Γ to be evidential of the "increasing power committees had to propose and dispose." This reads plausible, but it raises the question of how one can validate structural interpretations of discourse, especially when data is not intuitive—i.e., here both high and low resonance can be interpreted as influence, does it hurt construct validity?

[Twilight233333]

The author makes a profound study of speech in parliament and legislatures, including an analysis of the power to set the agenda and the novelty of speech. What I am interested in is whether it is possible to further track how an issue is transformed in the legislative, judicial and executive branches by including the political spectrum of the people in the executive branch, and if so, whether it is possible to take into account the media coverage of the time to observe how the issue is transformed between the people and the political institutions. On the left and the right, do they have an innate tendency that conservatives want to speak in parliament, while radicals prefer to have their views heard.

[joylin0209]

This study provides an in-depth analysis of the French Revolution debate, emphasizing three key constructs: novelty, transience, and resonance. I'm curious if the use of rhetoric or metaphor also brings a different impact to the debate? And, combining past research with modern politics, are there other factors influencing the novelty, transience, and resonance of speech in contemporary political debates, such as the rise of social media? (For example, how does social media interaction affect the resonance of a speech, and does this help build closer political connections?)

[michplunkett]

Regarding the current state of American politics, I found this observation of late 18th century French politics to be oddly familiar.

Indeed, the spatial metaphor of left and right is itself misleading: from the point of view of the debates themselves, the right wing appears as an inertial center, holding off conversation drift, while left-wing speakers produce a wide spectrum of innovations on a much larger periphery, only some of which survive.

Regarding the paper itself, I'm curious to what degree these observations are being applied to modern day elections and politicking. Knowledge for knowledge sake is certainly useful in aggregate, as most of the wonders of modern day life stem from investments in fundamental science research (think of the discoveries made during the mid-20th century space race), but I would be interested to see how papers like this and the one from last week regarding power relations and people changing their speech patterns could be utilized in modern day politicking. Are there political consultants that make use of these studies to coach and direct their candidates, or is the data coming from these large studies too high level at the moment to be directly applied to individuals' speech patterns?

[chanteriam]

The authors discuss their use of historical, French Revolution documents to examine how rhetorical and other strategies used reflected on following revolutions. As the way that language is used evolves over time, I wonder what considerations we should take when analyzing historical documents and applying it to current language patterns. Are there particular guidelines for using works that, for example, were written or expressed in Old English, studying their patterns, and applying them toward how we use language today? Or is the goal solely focused toward content, rather than syntax or semantics?

[yuzhouw313]

While this study utilized LDA to measure novelty, transience, and resonance within debates of the French Revolution, I wonder can we further deepen the topic analysis by incorporating network analysis. As Barron et al. highlighted the distinct language usage by left and right-wing orators, wonder can we construct a more nuanced hierarchical conversation network? Specifically, the speakers would serve as nodes, while the connections between them, formed through conversations and debates, would act as the edges. Such an approach could potentially reveal the intricate dynamics of influence and idea propagation among these historical figures, like the example in this week's codebook.

[donatellafelice]

This was a fascinating study! I am also curious about the framing and findings of this study. The authors wrote: "Taken together, these quantitative results align with existing qualitative interpretations, but also reveal crucial information-processing dynamics that have hitherto been overlooked" in the abstract. This raised a thought, in relation to both Text as Data and the Timmermans reading regarding the research process, does this mean that these types of methods should be run over all historical corpus that are available/can be created to confirm existing questions, the results of which could then be used to looked for new patterns? At what point are we publishing for novelty's sake, or looking for totally new questions, and what responsibility do we have to 'check' existing beliefs with these new methods? If you're trying to test someones qualitative analysis, you certainly need to be very cognizant to not end up doing the same inductive reasoning. Is it correct to use these methods to find gaps in the research, or we should be reevaluating the research all together?

[anzhichen1999]

In the study, resonance is defined as the difference between novelty and transience of speeches. How does this measure of resonance correlate with the long-term influence of these speeches in shaping public policy or social movements? Meanwhile, in regards of data density and brevity, is is challenging to apply the concept of resonance to analyze short, numerous texts like tweets? (after quick search it seems that it is not a common strategy)

[Caojie2001]

I think the paper's reasoning from the novelty–transience relationships to the emergence of the committee's new role is brilliant, as the change in the relationship pattern clearly demonstrates a new type of speech. However, I'm curious if the reasoning process shown in the article is consistent with the actual advancement of the study: is the division of committee and non-committee determined before analysis and supported by the analysis, or is it discovered after analysis to account for the change of pattern? Considering the actual research process, for those patterns found after primary analysis that existing theories can not explain, is there an effective way to figure out a meaningful dimension like the committee-non committee one in this research (I think it's an example of the abductive analysis mentioned in the orienting reading)?

[runlinw0525]

First, I think this exemplary reading is a very good example of using quantitative methods for social analysis. And considering the intriguing analysis of the French Revolution's National Constituent Assembly detailed in the paper, I'm curious: How do you think the distinct communication patterns and roles of left-wing and right-wing members, as well as the evolving influence of committees, shaped the overall dynamics and outcomes of the Revolution’s parliamentary debates?

[Brian-W00]

There are over 40,000 speeches used in this study, what were the specific challenges and limitations met in applying this methodology? And how did the researcher make sure the accuracy and reliability of the results given the historical and linguistic nuances of the period?

[naivetoad]

What implications does this study have for understanding the dynamics of modern parliamentary debates and political discourse? Can we identify similar patterns of linguistic innovation influenced by individual leaders or institutional changes in contemporary political discourse?

[chenyt16]

The study finds that the parliament favored the adoption of new patterns, with left-wing speakers innovating more and right-wing speakers preserving existing patterns. I am wondering if the linguistic innovations of left-wing speakers like Robespierre contributed to the shaping of revolutionary ideologies and policies. And whether this revolution only exists in political speech or also comes into people's daily lives. Can we discover a similar pattern with people obtaining different political preferences?

[Carolineyx]

I really appreciate how this paper examines the relationship between novelty, transience, and resonance, and the methodology it employs for their calculation and analysis. I'm curious—could we potentially develop a universal standard to predict the content that influences readers' or audiences' general sense of 'novelty'? If we avoid relying on previous and later comparisons and have a substantial amount of text, could we create a general model or prediction for readers' perceptions of novelty?

[ddlxdd]

I think the idea of “surprise” and “transience” are very interesting, And I think in linguistic, “surprise” is also a very important term to measure the change of speech in order to analyze the behavior change. So, I am wondering how to measures the surprise in the context of speech. I know it measures both the deviation of one speech from the patterns of prior ones (novelty)and from patterns that appear in the future, but how deviation is enough to be considered as novelty?

[JessicaCaishanghai]

Information theory has been employed to study the seemingly abysmal concepts like novelty, surprise, transience and resonance. Is this study replicable? Can we use this method to study all the other concepts' communications process?

[floriatea]

How do the evolving roles of committees, as identified in the early stages of the French Revolution, compare with the function and influence of committees in contemporary legislative bodies? Any lessons modern legislatures can draw from the revolutionary period regarding the balance between public debate and committee deliberation in fostering policy innovation?

[Vindmn1234]

How do the patterns of innovation and information processing identified in the French Revolution's parliamentary debates compare with those observed in contemporary legislative bodies? Do you see similar dynamics at play in terms of how new ideas are introduced, debated, and adopted or rejected, especially considering the impact of modern communication technologies and social media on public discourse and political ideology?