Overview of Muon POG tasks

Work in progress, under construction

This page shows an overview of muon POG tasks and the manpower situation

PPDCriticalUncoveredTasks

Short-term tasks

Tasks to be performed preferably until end of 2012. ALERT! Need heavy support from PAGs! Support Experts will provide help.

Reconstruction

Task Task type Available people Task Priority Comment GA
Implement a muon object independent of DT/CSC
  ? A result of global fit only with Tracker and RPC
Development M. Kim, D. Kovalskyi Normal DONE To be removed ? Is there any documentation twikis/talks to link from our pages?

Identification

Task Task type Available people Task Priority Comment GA
Implement standard Muon id selectors in CMSSW
  ? Create a selector for Loose, Soft, Tight and HighPt id
  ? Make a reference list of events to be used for crosschecks
Development M. Bachtis High   To be completed for 53X and 44X, priority was high, now it's still so. What is the reference list of events?
* Finish the tuning of the HighPt id*
  ? Study alternatives to nLayer>8 cut
Development P. Traczyk High PAG to help: EXO This is DONE, the Piotr recipes could be included in selectors/methods as well
* PF muon id*
  ? Fully document the current PF id decision tree
  ? Quantify the proportion of PF muons selected through different branches of the decision tree
    ? how many Global, Tracker and Standalone muons are selected depending on the event topology
  ? Study the differences between Tight muons and PF muons with tight quality cuts
    ? how many non-common events are selected, what are their particular features?
Development   High PAGs to help: any still keep this maybe reformulated ?
Study reconstruction and id efficiency for high pT muons
  ? Try using high mass DY events or boosted Zs
  ? Use cosmics data and MC
Development   High PAG to help: EXO see emails from Piotr about people interested. Kerstin/Aachen ?
Understand if we need a specific muon id for muon veto and jet disambiguation
Development   Normal PAG to help: SUS what does it mean exactly (not quite clear to me) ?
* Improve the method of extracting fake rates*
  ? Use fit parameters instead of counts
  ? Study random matching
  ? Understand features observed in 2012 data
Development M. Mooney Normal PAG to help: BPH propose priority becomes HIGH ? p/pt dependence of fake rates, dependence on the selection (K0, D0, B0)
Study the systematics of the Tag&Probe method
  ? Repeat checks done in MUO-10-004 with full statistics
  ? Understand how well we can trust that the efficiency measured @Z is valid for high pT
  ? Use alternative methods to measure id efficiency (e.g. cosmics)
  ? Study id efficiency vs distance to jets, vs. muon distance etc.
Development H. Brun Normal PAGs to help: any maintain this (maybe not with Hugues again), mention effs from orthogonal datasets and their comparison with T&P for syst. checks, methods for complex trigger paths (a-la-Dima)

Momentum assignment

Task Task type Available people Task Priority Comment GA
* PF muon momentum assignment*
  ? Fully document the current PF muon momentum assignment decision tree
  ? Propagate the uncertainty on pT measurement into PF muon
  ? Introduce high pt reconstructors as a part of PF momentum assignment
    ? study the performance of PF momentum assignment for high pt muons
  ? Quantify the proportion of PF muons with pT from different fits
    ? how often is Global, Tracker and Standalone pT used depending on the event topology
  ? Study the differences between reco::Muons pT and PF muon pT
    ? study in detail events with large difference in pT
  ? Improve the PF momentum assignment decision tree
    ? use benchmark plots to qualify performance
  ? study the correlation of post-cleaning and id tracker quality cuts
    ? compare the effect on far pT tails
Development M. Bachtis High PAGs to help: any it seems there is a good amount of repetitions here. Important is : what is really DONE by Mike and what is still to be finished ? Is it reasonable that Mike will provide also some documentations ?
* Finish the study of fully-segment based fit (in 52X) using APE’s*
  ? Use updated APE's from last muon alignment
Development A. Svyatkovskiy High PAG to help: EXO YES but unclear to me the time chaining of this and the related jobs

Momentum scale and resolution

Task Task type Available people Task Priority Comment GA
* Understand implications of muon alignment on high pt momentum resolution*
Development P. Traczyk High PAG to help: EXO Also Jordan ? people from I.Furic's group ?
* Rochester method for momentum scale correction*
  ? Finish systematic studies
  ? Apply the method to 2012 data
Development J. Han, A. Bodek High   what is it still missing ? only the recalibration after the 53X rereco ? Maybe priority is now normal ? Arie/JiYeon will do it anyway smile
* MuScleFit method at Z peak*
  ? Finish studies for 42X and 44X including MC and full acceptance
  ? Continue studies with 2012 data
Development S. Casasso, E. Migliore High   Same comment as above. status: Normal ? Stefano/Ernesto will do it anyway smile
MuScleFit method at J/psi peak
  ? Make studies with 2012 data
Development C. Dozen, G. Gokbulut, M. de Mattia Normal   this is OK ?
SIDRA method
  ? Finish studies for 42X and 44X
  ? Look at 2012 data
Development   Normal   Is there anybody interested from Juan's group ? Do anybody care to maintain this method in the future, or is it deemed not necessary ?
Understand non-Gaussian part of the resolution at intermediate pTs
  ? Study events where different reconstructors give different pT
  ? Use cosmics and use large statistics MC
Development P. Traczyk High PAG to help: EXO Not sure I know what is this. Piotr still working on this ? Interested PAG only EXO ?
* Endpoint method*
  ? Update general and endpoint studies with 2012 data
  ? Improve precision of cosmic endpoint
Development N. Kypreos High PAG to help: EXO I.Furic's group ? institutional resp ? we need the code and interim solutions ? PAG: EXO and B2G ?
Better understanding of momentum resolution at high pt
  ? Compare resolutions measured with collision and cosmic muons
  ? Measure resolutions with cosmic MC
  ? Develop methods to measure high pt resolution in endcaps
Development   Normal PAG to help: EXO only EXO or also B2G ? SUS ?

Isolation

Task Task type Available people Task Priority Comment GA
* Advanced methods of isolation measurement*
  ? Tune working points of cut based radial isolation
    ? study the PU dependence of different working points
  ? Simplify the MVA isolation algorithm
    ? study the use of dBeta PU correction instead of rho
  ? implement these algorithms in CMSSW
Development S. Folgueras High PAGs to help: HIG, any what is the status of this? and in CMSSW ? is Santiago really interested ? Groups other than HIG ? TOP, SUS, SMP ?
Quantify the systematics of isolation efficiency measurements
  ? Compare T&P and LKT
  ? Perform T&P syst. studies similar to those done for id in MUO-10-004
Development   Normal PAGS to help: SMP, any comparison of dBeta and EA corrections ? Again: help from HIG, TOP,SUS,SMP ? G.Ortona ?

Trigger

Task Task type Available people Task Priority Comment GA
* Understand and cure the L3 PU dependent inefficiency in 0.9<|eta|<1.5 region*
  ? Find the origin of the inefficiency (seeding, reconstruction)
  ? Cure the inefficiency (without increasing the rate) by varying relevant parameters
Development   High PAGs to help: any This is in the Trigger TSG list. Purdue signed on it ?
* L2 efficiency for close muons*
  ? Inefficiency due to too many shared hits
  ? Debug the seeding for close muons
  ? Understand why the reconstruction ends up with two overlapping tracks
Development F. Fiori High PAG to help: BPH talk to F.Fiori. Help from BPH or EXO/B2G ?
Isolation tuning
  ? Use PF isolation online
Development   Normal PAGs to help: any
Tuning of L2/L3 quality cuts
  ? Tune current and add new quality cuts (a la offline)
Development   Normal need to watch the effect on offline Reminder: what has been done in 2012 ? this has to go in parallel with the core developments
Quantify the systematics of trigger efficiency measurements
  ? Perform T&P syst. studies similar to those done in MUO-10-004
  ? Check T&P efficiencies with forcing the tag to be in different detector (link)
Development   Normal PAGS to help: any comparison of T&P and orthogonal datasets methods

Long-term tasks

The wish-list for developments in muon reconstruction and id during 2012 and LS1. This list is intended to collect ideas for improvements, as a complement to the Muon Reconstruction Workshop that has been held as part of the Muon Workshop in October.

  • Local Reconstruction
    • RPC
      • To include masked strips in DB to do a more precise reclusterization.
      • RPC Alignment.
      • RE4 in DQM and reconstruciton Local and Global.
      • Check error assignment in endcaps
    • DT
      • Full understanding of mean timer (in particular chi2 of STA in overlap region)
      • Introduce dead channels to simulation/reconstruction
      • Use muon RECO information on MB4 to correct time propagation along the wire
      • Study vdrift parameterization
      • Study possibility of applying quality cuts to segments in MB1 W+/-2
      • Combine all stations to produce a Super-z segment, could help to distinguish among different possibilities
      • Optimize hit errors given in input to the segment fit (now set to constant)
      • Study possibility of special treatment in case of showering muons
    • CSC
      • Ensure local reconstruction works correctly with LS1 upgrade hardware
      • Review parameters of UF rechit algorithm.
      • Review treatment of rechits and segments when showering occurs.
      • Remove crosstalk conditions data from rechit reconstruction.
      • Study influence of noise on rechit reconstruction.
      • General code cleanup for both CSCRecHitD and CSCSegAlgoST
      • Reintroduce possibility of using Gatti fit to obtain position rather than parameterization
      • Continue studies of rechit ‘charge’, and of rechit and segment time
      • Study possibility of using Hough-Legendre transformations

  • Standalone muon reconstruction
    • Improve efficiency for nearby muons (arbitration/seeding, see details in trigger part)
    • Understand differences between data and MC resolution (~10%) on 1/pt resolution
      • This difference has been observed in MUO-10-004 (Fig. 22)
      • According to studies of D. Trocino and N. Amapane it does not seem to depend on alignment scenario
    • Cleaning up the RecoMuon collection
      • This concerns duplicates entering the collection from different types of reconstruction (e.g. STA with or w/o update at vertex)
      • what is the status now ? are we sure what we want to do for muons which are ONLY STA ? in case of exotic signals like those searched by M.deMattia...
    • Study biases wrt. BS
      • For reference see latest studies of M. de Mattia here ok but maybe expand in view of the reconstruction of displaced muons ?
    • Understand the reason for the eta flip after update at vertex and fix it properly
      • For now only a protection is added when STA and STA updated at vertex have opposite eta sign in L2 there is a protection too !
      • For more details see talk by A. Everett here
      • link to Suneel's talks, minimal check of selected event lists to be done ?
    • No refit is currently applied for smoothing
      • It improves momentum resolution and provides correct track chi2
      • But found to be unstable and decrease efficiency, further studies needed
    • Study possibility of using the SIMPLEX method in standard STA. As it is not iterative it might be more stable than the kalman filter (for estimating pt, track parameter, estabilize the fit) No idea what it is ... who mentioned it ? references/links here ?
    • Try using RPC hits to confirm segments in 1st station
    • Study the idea of having tighter pattern recognition at the initial step (as done in the different iterative steps in inner tracking) maybe here the SET algorithm could be employed. We would need checks on possible performance...

  • Global muon reconstruction
    • Matching:
      • Reduce size of region of interest: improve use of vertex information
      • Revisit matching criteria to reduce number candidates per standalone muon (important for high pile-up and HLT)
      • Detailed study of the matching in the presence of high pile-up
      • Optimize algorithm for close-by muons
      • Study stability of local chi2 under misalignment
    • Use muon APEs (Alignment position errors) in STA and global fits
      • Pulls of global (re)fits at high pT suffer from large irregularities (see e.g. MUO-10-004 Fig. 23) which seems to be caused by underestimation of segment uncertainties in the fit; the inclusion of APE's seems to help curing the problem (see talk by J. Tucker here)
      • Studies are being done by A. Svyatkovskiy (see his presentations in muon POG e.g. here)
        • These studies are obscured by old MC muon misalignment scenario showing some coherent displacements (see next item)
      • Latest results by Alexey show good behavior of fit with APES when using a more realistic misalignment scenario (see here)
    • Low-pT
      • Refit subset of tracker muons with muon hits In general a full Kalman filtering In-OUT could be another approach ?? Gio.P. approach for specific tracking is doing something similar ?
    • High-pT
      • Investigate alternatives: try GSF, etc.
      • Exclude stations with hit multiplicity hit clusters (EM showers)
      • Consolidate current use of multiple refits
    • Understand better and improve the momentum scale and resolution for high pT muons
      • Recently it has been shown that the momentum scale and resolution strongly depends of the quality of the muon alignment (see talks by P. Traczyk, e.g. here)
        • All this needs deeper understanding in order to reach ultimate resolution at high pT according to TDR (9% mass resolution at 5 TeV mumu mass)
        • Latest explanation on the muon alignment in MC has been given here.
      • Momentum scale is studied using general and cosmics endpoint methods (see talks by N. Kypreos e.g. here)
        • The precision obtained is limited (~5%/TeV) due to low statistics. Precision of cosmics endpoint could be improved by carefully relaxing the selection criteria.
      • Resolution is determined using cosmics (see talks by J. Tucker e.g. here)
        • How much can we trust the resolution from cosmics - need to compare to collision data
    • Full use of DYT (Dynamic Truncation algorithm)
      • For more details see talk by D. Pagano here
      • This algorithm is the most complete for high pT momentum fitting but was not used in any analysis up to now
      • A prerequisite is the use of APEs
    • Study efficiency for high pT muons
      • For the moment we do not have a mature data-driven technique to study the efficiencies at high pT Orthogonal datasets ! on a clean selection of isolated muons
      • Some studies have been done in the past with cosmics (see talk by I. Belotelov here) but systematics specific to cosmics needs to be disentangled (new cosmics MC should be helpful)
      • Using full 2012 statistics one should be able to use collision data
    • Study the possible contribution of GEM on the global muon reconstruction
      • For details see e.g. talk of M. Maggi here
      • Work needs to be done to understand why the impact of ME1/1 is so small, observed in the above studies

    • Displaced Muons: complete the reco and fit trying the dedicated muon tracking
    • Check alignment workflow and the ultimate precision with the expected improvements (DT meantimer, DT proper error assignment, etc...)

  • Tracker Muons
    • Understand pull/residual differences between data and MC (~10%)
      • These differences have been observed e.g. in MUO-10-004 (sec. 4.3, Figs. 8 and 9)
      • This discrepancy strongly contributes to data/MC scale factors of Tracker Muon based muon id efficiency esp. in barrel (see Fig. 12 left in MUO-10-004)
    • Systematic comparison of selectors
    • Study of arbitration
      • in high-pileup environment (keeping low fake rate without cost in efficiency)
      • using RPC information for arbitration
    • Study further use of TM in HLT, even for single muon trigger. check what we are saying in other places (who ?)
    • Matching of Tracker to STA muons: useful and providing further power to reduce Fake rates (Bs Muon ID studies). Study more variables, code existing but almost neglected. DT improvements on the error assignment could boost these potential. Sensitivity to the alignment too. (people ? )
  • RPC Muons
    • Study of fake rates
    • Study of efficiencies
      • compare the efficiency to that from the standard Loose Muon selection
    • Study of track-to-hit matches
      • residual and pull
      • bunch crossing
    • Apply the RPC-standalone seed to track matching
      • RPC-standalone seed is a dedicated seed which uses only the RPC hits as input

  • Isolation
    • Further improvement of isolation algorithms
      • Improve uniformity vs. pT
        • The efficiency of currently used algorithms (all relative isolation) decreases at low pT; one possibility could be to combine the relative isolation with absolute isolation at low pT
      • Further improve uniformity vs. eta
        • First step done: here
      • Further improve uniformity vs. PU
      • Do we REALLY need to improve the uniformity ? It seems already pretty flat, at least the scale factors

  • GED (Global Event Description)
    • Study implications of post cleaning for different topologies
      • Important is to study the effect of post-cleaning on topologies with natural MET
        • Part of post cleaning is based on MET minimization (see talk by P. Janot here) Is this completely DONE by Mike ?
      • Study and improve high pT momentum assignment
        • See also short-term tasks
        • Incorporate global refits including DYT
    • Study the momentum assignment in dense environment
      • The most recent studies were focused on comparisons of tracker track and GED muon momentum at intermediate pT for isolated muons (talks by C. Battilana here and talk by C. Bernet here)
    • Further study and improve the recovery of momentum lost due to FSR (and EM showering)
      • Partially done in PF algo
      • Might need to use full event information - analysis specific (see talk by P. Janot here)

  • Tracking (together with TRK POG)
    • Outside-in seeding of muon tracks This is DONE by Giovanni P. !
      • After recent changes (52X) in tracking the average number of layers/hits used in the track fit has decreased for muons
      • The idea is to introduce a specific tracking for muons based on outside-in seeding using information from muon detectors
      • See here for latest developments
    • Further tracking development (MVA) using muon detectors Not looked into: what is it about ?
      • See ideas outlined in the talk by K. Stenson here

  • Trigger
    • Use Tracker Muon in trigger also for single muons Discuss with Dima. Who ?
    • Use segment fit at L2 (as in STA)
    • Try smoothing in refit (see here for potential improvements and this link for the associated efficiency drop CHCEK THIS !
    • Improve efficiency for nearby muons
      • Mainly L2, check also L3
      • See talks of F. Fiori and S. Zenz here CHECK S.ZENZ talk
        • Another talk by F. Fiori here shows that the problem might originate in L2(STA) seeding
    • Make L3 more robust against PU
      • Review current algorithms, develop new ones
      • Optimize L3 seeding
      • Re-optimize error matrix
      • Optimize L3 matching filters
      • Understand the PU dependent inefficiency in 0.9<|eta|<1.5 region (see also short-term tasks)
        • This has been shown to be reproduced by MC
        • For reference see plots by S. Walsh and B. Klein data, MC) CHECK this
    • More detailed studies of triggering on high PT muons
      • Use of global refits in trigger
    • Double muon trigger purity vs. PU What is it ?
      • We see increase of double muon trigger cross sections with PU which needs to be cleaned up without significant loss of efficiency
    • L2 purity as a function of L1 threshold What is it ?
      • We observe an increase of trigger rates depending of the L1 threshold (rate) which indicates that L1 does better job rejecting certain classes of background
  • Fast simulation
    • Various improvements: punch-trough, mu energy deposits in calorimeters, bremsstrahlung
  • Simulation * Cleaning of DigiSimLink classes definitions for CSC and RPC, Event Content for RECOSIM * Development of new TrackingParticles, used by both TRK and MUO POGs for validation. This is necessary for simulations for future High Pileup scenarios if one wants to keep our current muon validation, based on muon association by hits (Sim to Reco and viceversa). It has to be done jointly with the TRK community. (Urgent) * Development of a general container class to be used for all the tracking detectors to link the SimTracks/TrackingParticles to the RecHits. Now the TRK rechits have a class called OmniClusterRef which seem ok. We need to extend it for the muon (DT, CSC, RPC) RecHit formats. (urgent)

Continuous tasks

Muon reconstruction and identification

Task Task type Available people Task Priority Comment GA
Software coordination: integration, validation and publication of changes Continuous M. Bachtis, R. Bellan Normal  
Global muon reconstruction: maintenance, debugging, bug fixes On-demand   Normal Expert urgently needed!! Purdue ?
Standalone muon reconstruction: maintenance, debugging, bug fixes On-demand D. Trocino Normal Need 2nd expert for continuity ????
Tracker muon reconstruction: maintenance, debugging, bug fixes On-demand D. Kovalskyi Normal Need 2nd expert for continuity Dima/Claudio: proposals ?

Muon reco+id performance

Task Task type Available people Task Priority Comment GA
Measure efficiency of standard muon selections with T&P on J/ψ and Z peaks On-demand Go fast forward I. Kratschmer, I. Josa Mutuberria, E. Navarro, C. Botta, G. Petrucciani High  
Measure fake rates of standard muon selections with exclusive samples of π, K and p On-demand M. Mooney Normal   HIGH ? Include other studies for Bs (D0, B0) (Danek K.) and studies of improved MuonId for Bs (Galanti, Pazzini,..)
Measure reco/id efficiency for high-pT muons using cosmics data and MC On-demand/Development   Normal a way to validate MC-based high-pT muon efficiency
Go fast forward - short learning phase (no/little previous expertise required)

Muon isolation

Task Task type Available people Task Priority Comment GA
Measure reference isolation efficiencies On-demand Go fast forward I. Kratschmer, I. Josa Mutuberria, E. Navarro Normal   G.Ortona ? Santiago ?
Go fast forward - short learning phase (no/little previous expertise required)

Muon momentum scale & resolution

Task Task type Available people Task Priority Comment GA
Calibration of p scale and measurement of the p resolution On-demand J. Han, A. Bodek, S. Casasso, E. Migliore High See short-term tasks
Evaluate momentum scale at high pT (cosmic endpoint method) On-demand N. Kypreos Normal   No more Nick. I.Furic's group ?
Measure resolution of various reconstructors at high pT using cosmics On-demand   Normal need a person to replace J. Tucker or Jordan again ? talk him ?

Certification/Validation/DQM

Task Task type Available people Task Priority Comment GA
Release validation with MC Continuous R. Vilar, A. Perotta, P. Martinez (hlt), A. George (id) High   remove Andrea and Pablo, here we need 1 person for HLT validation plus one other to help Rocio (some developments of the validation/DQM has to be foreseen as well).
Data certification; release validation on data; offline DQM Continuous A. Calderon Tazon, C. Jorda Lope, S. Folgueras High   Note that also Giovanni P. and Cristina, and Hugues Brun have helped on this!
Improve DQM: improve PU monitoring, put (cosmic) high pt resolution into DQM, muon alignment validation, add more selectors, improve GED validation tool for muons Development A. Calderon Tazon, S. Folgueras Normal need additional people Clara has contributed already here, with Giovanni P. too.
Automate certification process Development   Normal Important to avoid oversights and release load on experts Note that Giovanni P. has helped with the script to validate new conditions/rereco on data with a T&P without fitting
Automate validation process Development   Low Ditto

Tools

Task Task type Available people Task Priority Comment GA
Cosmic identification tool development Development H. Yoo Low  
Beam-halo identification tool development Development R. Remington Low  
T&P package: maintenance and bug fixes On-demand G. Petrucciani, C. Botta Normal Need more experts for continuity
T&P package: development Development G. Petrucciani, C. Botta Normal  
MuScleFit maintenance On-demand E. Migliore, S. Casasso Normal  
Rochester corrections maintenance On-demand A.Bodek, J.Han Normal  
SIDRA maintenance On-demand   Low  
Kink-finder: maintenance and bug fixes On-demand G. Petrucciani Low  
Shower id variables: further studies Development A. Svyatkovskiy Low  
MC Truth tool (MuonAssociatorByHits) On-Demand G.Abbiendi Normal

HLT

Task Task type Available people Task Priority Comment GA
Coordination, menu survey Permanent D. Trocino High Would need another person to join Daniele
Level-3: Maintenance, debugging, bug fixes On-demand   High need an expert
Level-2: Maintenance, debugging, bug fixes On-demand D. Trocino High Same person as STA; need 2nd expert
Trigger efficiency measurements with T&P On-demand Go fast forward S. Walsh, B. Klein, O. Hindrichs, I. Kratschmer, I. Josa Mutuberria, E. Navarro Normal  
Trigger purity measurements On-demand Go fast forward Y. Choi, Y. Cho Normal  
HLT validation On-demand P. Martinez High   Pablo is gone, need one person
Improvements in HLT validation and DQM Development P. Martinez Normal   Pablo is gone, need one person
Trigger isolation efficiency measurements On-demand Go fast forward M. Zeise Normal  
Go fast forward - short learning phase (no/little previous expertise required)

Other

Task Task type Available people Task Priority Comment GA
Production of MC samples On-demand G. Abbiendi Normal Need 2nd expert for contingency a person is needed (even shared with Tracking POG)
Data transfer and management of group space at dedicated muon Tier-2's On-demand A. Fanfani Normal  
Fast simulation: punch-through outside tracker, μ energy dep. in calorimeters, bremsstrahlung Development A. Perotta Low  
| Simulation infrastructure developments (TrackingParticle reengineering et al.) | Development | | High | G.Abbiendi following up with Tracking POG people and Vincenzo Innocente, technical help could be needed | ??? |

Contacts

Review status

Reviewer/Editor and Date (copy from screen) Comments

%RESPONSIBLE% MuonPOG
%REVIEW% Most recent reviewer

Edit | Attach | Watch | Print version | History: r5 < r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r5 - 2014-05-05 - GiovanniAbbiendi
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Sandbox All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright &© 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback